Use Cases Archives

Attribution Modelling using Data Science and Deep Learning

Data-driven Attribution Modeling

October 20, 2024/in Artificial Intelligence, Business Analytics, Business Intelligence, Data Science, Deep Learning, Machine Learning, Main Category, Use Cases/by Alexander Lammers

In the world of commerce, companies often face the temptation to reduce their marketing spending, especially during times of economic uncertainty or when planning to cut costs. However, this short-term strategy can lead to long-term consequences that may hinder a company’s growth and competitiveness in the market.

Maintaining a consistent marketing presence is crucial for businesses, as it helps to keep the company at the forefront of their target audience’s minds. By reducing marketing efforts, companies risk losing visibility and brand awareness among potential clients, which can be difficult and expensive to regain later. Moreover, a strong marketing strategy is essential for building trust and credibility with prospective customers, as it demonstrates the company’s expertise, values, and commitment to their industry.

Given a fixed budget, companies apply economic principles for marketing efforts and need to spend a given marketing budget as efficient as possible. In this view, attribution models are an essential tool for companies to understand the effectiveness of their marketing efforts and optimize their strategies for maximum return on investments (ROI). By assigning optimal credit to various touchpoints in the customer journey, these models provide valuable insights into which channels, campaigns, and interactions have the greatest impact on driving conversions and therefore revenue. Identifying the most important channels enables companies to distribute the given budget accordingly in an optimal way.

1. Combining business value with attribution modeling

The true value of attribution modeling lies not solely in applying the optimal theoretical concept – that are discussed below – but in the practical application in coherence with the business logic of the firm. Therefore, the correct modeling ensures that companies are not only distributing their budget in an optimal way but also that they incorporate the business logic to focus on an optimal long-term growth strategy.

Understanding and incorporating business logic into attribution models is the critical step that is often overlooked or poorly understood. However, it is the key to unlocking the full potential of attribution modeling and making data-driven decisions that align with business goals. Without properly integrating the business logic, even the most sophisticated attribution models will fail to provide actionable insights and may lead to misguided marketing strategies.

Figure 1 – Combining the business logic with attribution modeling to generate value for firms

For example, determining the end of a customer journey is a critical step in attribution modeling. When there are long gaps between customer interactions and touchpoints, analysts must carefully examine the data to decide if the current journey has concluded or is still ongoing. To make this determination, they need to consider the length of the gap in relation to typical journey durations and assess whether the gap follows a common sequence of touchpoints. By analyzing this data in an appropriate way, businesses can more accurately assess the impact of their marketing efforts and avoid attributing credit to touchpoints that are no longer relevant.

Another important consideration is accounting for conversions that ultimately lead to returns or cancellations. While it’s easy to get excited about the number of conversions generated by marketing campaigns, it’s essential to recognize that not all conversions should be valued equal. If a significant portion of conversions result in returns or cancellations, the true value of those campaigns may be much lower than initially believed.

To effectively incorporate these factors into attribution models, businesses need to important things. First, a robust data platform (such as a customer data platform; CDP) that can integrate data from various sources, such as tracking systems, ERP systems, e-commerce platforms to effectively perform data analytics. This allows for a holistic view of the customer journey, including post-conversion events like returns and cancellations, which are crucial for accurate attribution modeling. Second, as outlined above, businesses need a profound understanding of the business model and logic.

2. On the Relevance of Attribution Models in Online Marketing

A conversion is a point in the customer journey where a recipient of a marketing message performs a somewhat desired action. For example, open an email, click on a call-to-action link or go to a landing page and fill out a registration. Finally, the ultimate conversion would be of course buying the product. Attribution models serve as frameworks that help marketers assess the business impact of different channels on a customer’s decision to convert along a customer´s journey. By providing insights into which interactions most effectively drive sales, these models enable more efficient resource allocation given a fixed budget.

Figure 2 – A simple illustration of one single customer journey. Consider that from the company’s perspective all journeys together result into a complex network of possible journey steps.

Companies typically utilize a diverse marketing mix, including email marketing, search engine advertising (SEA), search engine optimization (SEO), affiliate marketing, and social media. Attribution models facilitate the analysis of customer interactions across these touchpoints, offering a comprehensive view of the customer journey.

Comprehensive Customer Insights: By identifying the most effective channels for driving conversions, attribution models allow marketers to tailor strategies that enhance customer engagement and improve conversion rates.
Optimized Budget Allocation: These models reveal the performance of various marketing channels, helping marketers allocate budgets more efficiently. This ensures that resources are directed towards channels that offer the highest return on investment (ROI), maximizing marketing impact.
Data-Driven Decision Making: Attribution models empower marketers to make informed, data-driven decisions, leading to more effective campaign strategies and better alignment between marketing and sales efforts.

In the realm of online advertising, evaluating media effectiveness is a critical component of the decision-making process. Since advertisement costs often depend on clicks or impressions, understanding each channel’s effectiveness is vital. A multi-channel attribution model is necessary to grasp the marketing impact of each channel and the overall effectiveness of online marketing activities. This approach ensures optimal budget allocation, enhances ROI, and drives successful marketing outcomes.

What types of attribution models are there? Depending on the attribution model, different values are assigned to various touchpoints. These models help determine which channels are the most important and should be prioritized. Each channel is assigned a monetary value based on its contribution to success. This weighting then determines the allocation of the marketing budget. Below are some attribution models commonly used in marketing practice.

2.1. Single-Touch Attribution Models

As it follows from the name of the group of these approaches, they consider only one touchpoint.

2.1.1 First Touch Attribution

First touch attribution is the standard and simplest method for attributing conversions, as it assigns full credit to the first interaction. One of its main advantages is its simplicity; it is a straightforward and easy-to-understand approach. Additionally, it allows for quick implementation without the need for complex calculations or data analysis, making it a convenient choice for organizations looking for a simple attribution method. This model can be particularly beneficial when the focus is solely on demand generation. However, there are notable drawbacks to first touch attribution. It tends to oversimplify the customer journey by ignoring the influence of subsequent touchpoints. This can lead to a limited view of channel performance, as it may disproportionately credit channels that are more likely to be the first point of contact, potentially overlooking the contributions of other channels that assist in conversions.

Figure 3 – The first touch is a simple non-intelligent way of attribution.

2.1.2 Last Touch Attribution

Last touch attribution is another straightforward method for attributing conversions, serving as the opposite of first touch attribution by assigning full credit to the last interaction. Its simplicity is one of its main advantages, as it is easy to understand and implement without the need for complex calculations or data analysis. This makes it a convenient choice for organizations seeking a simple attribution approach, especially when the focus is solely on driving conversions. However, last touch attribution also has its drawbacks. It tends to oversimplify the customer journey by neglecting the influence of earlier touchpoints. This approach provides limited insights into the full customer journey, as it focuses solely on the last touchpoint and overlooks the cumulative impact of multiple touchpoints, missing out on valuable insights.

Figure 4 – Last touch attribution is the counterpart to the first touch approach.

2.2 Multi-Touch Attribution Models

We noted that single-touch attribution models are easy to interpret and implement. However, these methods often fall short in assigning credit, as they apply rules arbitrarily and fail to accurately gauge the contribution of each touchpoint in the consumer journey. As a result, marketers may make decisions based on skewed data. In contrast, multi-touch attribution leverages individual user-level data from various channels. It calculates and assigns credit to the marketing touchpoints that have influenced a desired business outcome for a specific key performance indicator (KPI) event.

2.2.1 Linear Attribution

Linear attribution is a standard approach that improves upon single-touch models by considering all interactions and assigning them equal weight. For instance, if there are five touchpoints in a customer’s journey, each would receive 20% of the credit for the conversion. This method offers several advantages. Firstly, it ensures equal distribution of credit across all touchpoints, providing a balanced representation of each touchpoint’s contribution to conversions. This approach promotes fairness by avoiding the overemphasis or neglect of specific touchpoints, ensuring that credit is distributed evenly among channels. Additionally, linear attribution is easy to implement, requiring no complex calculations or data analysis, which makes it a convenient choice for organizations seeking a straightforward attribution method. However, linear attribution also has its drawbacks. One significant limitation is its lack of differentiation, as it assigns equal credit to each touchpoint regardless of their actual impact on driving conversions. This can lead to an inaccurate representation of the effectiveness of individual touchpoints. Furthermore, linear attribution ignores the concept of time decay, meaning it does not account for the diminishing influence of earlier touchpoints over time. It treats all touchpoints equally, regardless of their temporal proximity to the conversion event, potentially overlooking the greater impact of more recent interactions.

Figure 5 – Linear uniform attribution.

2.2.2 Position-based Attribution (U-Shaped Attribution & W-Shaped Attribution)

Position-based attribution, encompassing both U-shaped and W-shaped models, focuses on assigning the most significant weight to the first and last touchpoints in a customer’s journey. In the W-shaped attribution model, the middle touchpoint also receives a substantial amount of credit. This approach offers several advantages. One of the primary benefits is the weighted credit system, which assigns more credit to key touchpoints such as the first and last interactions, and sometimes additional key touchpoints in between. This allows marketers to highlight the importance of these critical interactions in driving conversions. Additionally, position-based attribution provides flexibility, enabling businesses to customize and adjust the distribution of credit according to their specific objectives and customer behavior patterns. However, there are some drawbacks to consider. Position-based attribution involves a degree of subjectivity, as determining the specific weights for different touchpoints requires subjective decision-making. The choice of weights can vary across organizations and may affect the accuracy of the attribution results. Furthermore, this model has limited adaptability, as it may not fully capture the nuances of every customer journey, given its focus on specific positions or touchpoints.

Figure 6 – The U-shaped attribution (sometimes known as “bathtube model” and the W-shaped one are first attempts of weighted models.

2.2.3 Time Decay Attribution

Time decay attribution is a model that primarily assigns most of the credit to interactions that occur closest to the point of conversion. This approach has several advantages. One of its key benefits is temporal sensitivity, as it recognizes the diminishing impact of earlier touchpoints over time. By assigning more credit to touchpoints closer to the conversion event, it reflects the higher influence of recent interactions. Additionally, time decay attribution offers flexibility, allowing organizations to customize the decay rate or function. This enables businesses to fine-tune the model according to their specific needs and customer behavior patterns, which can be particularly useful for fast-moving consumer goods (FMCG) companies. However, time decay attribution also has its drawbacks. One challenge is the arbitrary nature of the decay function, as determining the appropriate decay rate is both challenging and subjective. There is no universally optimal decay function, and choosing an inappropriate model can lead to inaccurate credit distribution. Moreover, this approach may oversimplify time dynamics by assuming a linear or exponential decay pattern, which might not fully capture the complex temporal dynamics of customer behavior. Additionally, time decay attribution primarily focuses on the temporal aspect and may overlook other contextual factors that influence touchpoint effectiveness, such as channel interactions, customer segments, or campaign-specific dynamics.

Figure 7 – Time-based models can be configurated by according to the first or last touch and weighted by the timespan in between of each touchpoint.

2.3 Data-Driven Attribution Models

2.3.1 Markov Chain Attribution

Markov chain attribution is a data-driven method that analyzes marketing effectiveness using the principles of Markov Chains. Those chains are mathematical models used to describe systems that transition from one state to another in a chain-like process. The principles focus on the transition matrix, derived from analyzing customer journeys from initial touchpoints to conversion or no conversion, to capture the sequential nature of interactions and understand how each touchpoint influences the final decision. Let’s have a look at the following simple example with three channels that are chained together and leading to either a conversion or no conversion.

Figure 8 – Example of four customer journeys

The model calculates the conversion likelihood by examining transitions between touchpoints. Those transitions are depicted in the following probability tree.

Figure 9 – Example of a touchpoint network based on customer journeys

Based on this tree, the transition matrix can be constructed that reveals the influence of each touchpoint and thus the significance of each channel.

This method considers the sequential nature of customer journeys and relies on historical data to estimate transition probabilities, capturing the empirical behavior of customers. It offers flexibility by allowing customization to incorporate factors like time decay, channel interactions, and different attribution rules.

Markov chain attribution can be extended to higher-order chains, where the probability of transition depends on multiple previous states, providing a more nuanced analysis of customer behavior. To do so, the Markov process introduces a memory parameter 0 that is assumed to be zero here. Overall, it offers a robust framework for understanding the influence of different marketing touchpoints.

2.3.2 Shapley Value Attribution (Game Theoretical Approach)

The Shapley value is a concept from game theory that provides a fair method for distributing rewards among participants in a coalition. It ensures that both gains and costs are allocated equitably among actors, making it particularly useful when individual contributions vary but collective efforts lead to a shared outcome. In advertising, the Shapley method treats the advertising channels as players in a cooperative game. Now, consider a channel coalition consisting of different advertising channels . The utility function describes the contribution of a coalition of channels .

In this formula, is the cardinality of a specific coalition and the sum extends over all subsets of that do not contain the marginal contribution of channel to the coalition . For more information on how to calculate the marginal distribution, see Zhao et al. (2018).

The Shapley value approach ensures a fair allocation of credit to each touchpoint based on its contribution to the conversion process. This method encourages cooperation among channels, fostering a collaborative approach to achieving marketing goals. By accurately assessing the contribution of each channel, marketers can gain valuable insights into the performance of their marketing efforts, leading to more informed decision-making. Despite its advantages, the Shapley value method has some limitations. The method can be sensitive to the order in which touchpoints are considered, potentially leading to variations in results depending on the sequence of attribution. This sensitivity can impact the consistency of the outcomes. Finally, Shapley value and Markov chain attribution can also be combined using an ensemble attribution model to further reduce the generalization error (Gaur & Bharti 2020).

2.33. Algorithmic Attribution using binary Classifier and (causal) Machine Learning

While customer journey data often suffices for evaluating channel contributions and strategy formulation, it may not always be comprehensive enough. Fortunately, companies frequently possess a wealth of additional data that can be leveraged to enhance attribution accuracy by using a variety of analytics data from various vendors. For examples, companies might collect extensive data, including customer website activity such as clicks, page views, and conversions. This data includes features like for example the Urchin Tracking Module (UTM) information such as source, medium, campaign, content and term as well as campaign, device type, geographical information, number of user engagements, and scroll frequency, among others.

Utilizing this information, a binary classification model can be trained to predict the probability of conversion at each step of the multi touch attribution (MTA) model. This approach not only identifies the most effective channels for conversions but also highlights overvalued channels. Common algorithms include logistic regressions to easily predict the probability of conversion based on various features. Gradient boosting also provides a popular ensemble technique that is often used for unbalanced data, which is quite common in attribution data. Moreover, random forest models as well as support vector machines (SVMs) are also frequently applied. When it comes to deep learning models, that are often used for more complex problems and sequential data, Long Short-Term Memory (LSTM) networks or Transformers are applied. Those models can capture the long-range dependencies among multiple touchpoints.

Figure 10 – Attribution Model based on Deep Learning / AI

The approach is scalable, capable of handling large volumes of data, making it ideal for organizations with extensive marketing campaigns and complex customer journeys. By leveraging advanced algorithms, it offers more accurate attribution of credit to different touchpoints, enabling marketers to make informed, data-driven decisions.

All those models are part of the Machine Learning & AI Toolkit for assessing MTA. And since the business world is evolving quickly, newer methods such as double Machine Learning or causal forest models that are discussed in the marketing literature (e.g. Langen & Huber 2023) in combination with eXplainable Artificial Intelligence (XAI) can also be applied as well in the DATANOMIQ Machine Learning and AI framework.

3. Conclusion

As digital marketing continues to evolve in the age of AI, attribution models remain crucial for understanding the complex customer journey and optimizing marketing strategies. These models not only aid in effective budget allocation but also provide a comprehensive view of how different channels contribute to conversions. With advancements in technology, particularly the shift towards data-driven and multi-touch attribution models, marketers are better equipped to make informed decisions that enhance quick return on investment (ROI) and maintain competitiveness in the digital landscape.

Several trends are shaping the evolution of attribution models. The increasing use of machine learning in marketing attribution allows for more precise and predictive analytics, which can anticipate customer behavior and optimize marketing efforts accordingly. Additionally, as privacy regulations become more stringent, there is a growing focus on data quality and ethical data usage (Ethical AI), ensuring that attribution models are both effective and compliant. Furthermore, the integration of view-through attribution, which considers the impact of ad impressions that do not result in immediate clicks, provides a more holistic understanding of customer interactions across channels. As these models become more sophisticated, they will likely incorporate a wider array of data points, offering deeper insights into the customer journey.

Unlock your marketing potential with a strategy session with our DATANOMIQ experts. Discover how our solutions can elevate your media-mix models and boost your organization by making smarter, data-driven decisions.

References

Zhao, K., Mahboobi, S. H., & Bagheri, S. R. (2018). Shapley value methods for attribution modeling in online advertising. arXiv preprint arXiv:1804.05327.
Gaur, J., & Bharti, K. (2020). Attribution modelling in marketing: Literature review and research agenda. Academy of Marketing Studies Journal, 24(4), 1-21.
Langen H, Huber M (2023) How causal machine learning can leverage marketing strategies: Assessing and improving the performance of a coupon campaign. PLoS ONE 18(1): e0278937. https://doi.org/10.1371/journal. pone.0278937

DATANOMIQ Cloud Architecture for Data Mesh - Process Mining, BI and Data Science Applications

Data Mesh Architecture on Cloud for BI, Data Science and Process Mining

July 23, 2023/in Artificial Intelligence, Big Data, Business Analytics, Business Intelligence, Cloud, Data Engineering, Data Science, Machine Learning, Main Category, Predictive Analytics, Process Mining, Tool Introduction, Use Cases/by Benjamin Aunkofer

Companies use Business Intelligence (BI), Data Science, and Process Mining to leverage data for better decision-making, improve operational efficiency, and gain a competitive edge. BI provides real-time data analysis and performance monitoring, while Data Science enables a deep dive into dependencies in data with data mining and automates decision making with predictive analytics and personalized customer experiences. Process Mining offers process transparency, compliance insights, and process optimization. The integration of these technologies helps companies harness data for growth and efficiency.

Applications of BI, Data Science and Process Mining grow together

More and more all these disciplines are growing together as they need to be combined in order to get the best insights. So while Process Mining can be seen as a subpart of BI while both are using Machine Learning for better analytical results. Furthermore all theses analytical methods need more or less the same data sources and even the same datasets again and again.

Bring separate(d) applications together with Data Mesh

While all these analytical concepts grow together, they are often still seen as separated applications. There often remains the question of responsibility in a big organization. If this responsibility is decided as not being a central one, Data Mesh could be a solution.

Data Mesh is an architectural approach for managing data within organizations. It advocates decentralizing data ownership to domain-oriented teams. Each team becomes responsible for its Data Products, and a self-serve data infrastructure is established. This enables scalability, agility, and improved data quality while promoting data democratization.

In the context of a Data Mesh, a Data Product refers to a valuable dataset or data service that is managed and owned by a specific domain-oriented team within an organization. It is one of the key concepts in the Data Mesh architecture, where data ownership and responsibility are distributed across domain teams rather than centralized in a single data team.

A Data Product can take various forms, depending on the domain’s requirements and the data it manages. It could be a curated dataset, a machine learning model, an API that exposes data, a real-time data stream, a data visualization dashboard, or any other data-related asset that provides value to the organization.

However, successful implementation requires addressing cultural, governance, and technological aspects. One of this aspect is the cloud architecture for the realization of Data Mesh.

Example of a Data Mesh on Microsoft Azure Cloud using Databricks

The following image shows an example of a Data Mesh created and managed by DATANOMIQ for an organization which uses and re-uses datasets from various data sources (ERP, CRM, DMS, IoT,..) in order to provide the data as well as suitable data models as data products to applications of Data Science, Process Mining (Celonis, UiPath, Signavio & more) and Business Intelligence (Tableau, Power BI, Qlik & more).

Data Mesh on Azure Cloud with Databricks and Delta Lake for Applications of Business Intelligence, Data Science and Process Mining.

Microsoft Azure Cloud is favored by many companies, especially for European industrial companies, due to its scalability, flexibility, and industry-specific solutions. It offers robust IoT and edge computing capabilities, advanced data analytics, and AI services. Azure’s strong focus on security, compliance, and global presence, along with hybrid cloud capabilities and cost management tools, make it an ideal choice for industrial firms seeking to modernize, innovate, and improve efficiency. However, this concept on the Azure Cloud is just an example and can easily be implemented on the Google Cloud (GCP), Amazon Cloud (AWS) and now even on the SAP Cloud (Datasphere) using Databricks.

Databricks is an ideal tool for realizing a Data Mesh due to its unified data platform, scalability, and performance. It enables data collaboration and sharing, supports Delta Lake for data quality, and ensures robust data governance and security. With real-time analytics, machine learning integration, and data visualization capabilities, Databricks facilitates the implementation of a decentralized, domain-oriented data architecture we need for Data Mesh.

Furthermore there are also alternate architectures without Databricks but more cloud-specific resources possible, for Microsoft Azure e.g. using Azure Synapse instead. See this as an example which has many possible alternatives.

Summary – What value can you expect?

With the concept of Data Mesh you will be able to access all your organizational internal and external data sources once and provides the data as several data models for all your analytical applications. The data models are seen as data products with defined value, costs and ownership. Each applications has its own data model. While Data Science Applications have more raw data, BI applications get their well prepared star schema galaxy models, and Process Mining apps get normalized event logs. Using data sharing (in Databricks: Delta Sharing) data products or single datasets can be shared through applications and owners.

The Role Data Plays in HR Analytics

May 1, 2023/in Insights, Use Cases/by Shannon Flynn

Data analytics in HR can help businesses make informed decisions for hiring, promotion and digital transformation. While human resources is typically considered a “soft” discipline, information can reveal invaluable insights that help professionals deliver tangible improvements. What role does data play in HR analytics and success?

The Value of Data in HR

Data is crucial for the success of HR analytics tools. It increases visibility into business processes and the employee experience. The information analytics reveals allows decisions to be based on proven facts rather than subjective assumptions.

For example, professional absenteeism costs an estimated $24.2 billion annually worldwide. Reducing these rates relies on identifying the most common causes of missing work among employees. Some might have a chronic illness or unpredictable family obligations. Others might struggle to maintain motivation if the workplace culture does not fit them well.

Data highlights information like this, allowing HR professionals to act on sound evidence and insights. This applies to HR-specific choices, like hiring, as well as businesswide decisions, like the best way to implement a new app or technology.

Applications for HR Analytics Tools

What are the benefits of data in HR? There are many applications for the insights gained from HR analytics tools. Most business goals and challenges are connected to HR in one way or another, so data-powered solutions can have a significant ripple effect.

Data-Driven Hiring

Refining the recruitment process is a top priority for many HR professionals. Data analytics can streamline hiring, from finding potential candidates to choosing new hires.

Important KPIs for this category include time to hire, time to fill, offer acceptance rates and application sources. This data highlights how applicants hear about the business’s job openings, how long it takes to fill open positions and how frequently first-choice applicants accept job offers. Analyzing these key data types lets HR professionals pinpoint ways to improve their hiring process.

For example, HR analytics tools could reveal that a certain job board is more likely to attract applicants who accept offers. It might refer fewer candidates than another, but data would show it attracts higher-quality candidates. HR managers could then focus on prioritizing their postings on that specific site.

More Informed Employee Management

Data analytics in HR enable employee management decisions, like promotions, to be based on hard numerical data. This can be particularly helpful since getting or missing a promotion can affect workers emotionally. If they can see why they may not have received a promotion, they may be more likely to turn disappointment into motivation to improve.

HR professionals can track KPIs like average projects completed each month, client reviews, performance over time or job review results. Analyzing all this data can highlight employees who may fly under the radar while actually outperforming colleagues. Insights like this allow HR professionals to make more informed promotion decisions.

Digital Transformation Initiatives

Employees play a core role in the success of digital transformation. Surveys show that 27% of business executives are concerned about where to focus their efforts. Applying data in HR can reveal employees’ needs and the areas where technology upgrades can best serve them.

HR professionals can also use data analytics to measure the success of digital transformation initiatives once they are implemented. For example, employee performance and satisfaction surveys might show most workers like a new software program but find it confusing to learn. The HR department could use this information to suggest more thorough training for new tools moving forward.

Artificial Intelligence for HR

Data analytics in HR doesn’t need to be a manual process. Cutting-edge AI tools are now widely available to help with the data analysis process. For example, ChatGPT, one of today’s most popular AI models, can give users instructions on how to set up data analytics programs.

While ChatGPT won’t replace professional data analysts any time soon, it can help HR professionals navigate analytics. An HR department might want a data analysis program to predict employee success. It can use ChatGPT to generate instructions on creating that specific program. This technology can even write functional code.

Artificial intelligence for HR data analysis is still somewhat limited. ChatGPT may be smart but can only work with text input and output. However, AI does make a helpful assistant. Additionally, collecting large amounts of information is central to creating and training well-optimized AI models. Compiling HR data can help prepare businesses for the future.

Data-Driven Human Relations

Data analytics and artificial intelligence for HR can revolutionize decision-making. HR analytics tools ground promotions and hiring in clear numerical KPIs. Businesses can even use data to analyze the performance of big changes, like integrating a new digital transformation initiative. It opens the door to many possibilities that can lead to an even more productive human resources department.

Data-Driven Approaches to Improve Senior Living

April 13, 2023/in Use Cases/by Shannon Flynn

Data-driven approaches have become standard across many industries, but some still need to catch up for using Big Data. Health care has slowly embraced digital transformation and data analytics, but senior living facilities have room to improve under that umbrella. If more long-term care (LTC) organizations adopted data initiatives, they could significantly improve their patients’ standards of living.

Almost a quarter of LTC providers report having “very little” ability to access and share patient data electronically. Nearly a third still rely on email or fax for these processes, and 18% are entirely manual. Consequently, data analytics remains an area of untapped potential for many of these facilities.

Personalizing Care

Individualized care is one of the most promising applications of data analytics in health care and senior living is no exception. Using machine learning to analyze electronic health records would enable LTC organizations to tailor care to individual patients.

AI can analyze patients’ medical history and larger trends among similar cases to determine what steps may result in the best health outcomes for each patient. Personalized plans of care like this have yielded favorable results, such as 12% reductions in emergency room visits and 8% increases in medication adherence.

As more LTC facilities use data analytics to personalize care, they’ll generate more data on which steps work best for different cases. This data will lead to long-term improvements, making AI an increasingly reliable personalization tool.

Accelerating Emergency Response

Data-driven approaches can also help senior living facilities respond faster to any emergencies. Wearables and other Internet of Things technologies can track health factors like heart rates, body temperatures and more, analyzing this data in real-time to monitor for abnormalities. As soon as anything falls out of acceptable parameters, they can alert medical staff.

AI can often detect trends in data and interpret signals earlier and more accurately than humans. As a result, these early warnings could lead to unprecedented improvements in emergency response times, significantly improving patient outcomes.

In a 2022 study, 86% of patients agreed their health-monitoring wearables improved their health and quality of life. Even without emergencies, these results suggest implementing data-centric technologies can improve standards of living and satisfaction in LTC.

Streamlining Operations

LTC organizations can also achieve less critical but still essential benefits from data initiatives. Transitioning from paper and manual processes to embrace electronic data and automation will boost organizational efficiency and lower costs.

Analyzing data on workflows like response times, patient surveys, incident numbers and similar information can reveal where organizations can do better and where they’re doing well. These insights, in turn, guide more effective decision-making on reorganizing workflows or editing policies to improve standards of living or reduce costs.

As LTC organizations become more cost-efficient, they can lower patient costs. Those savings are crucial, considering 90% of American adults don’t have long-term care insurance, despite more than half needing such care. Using data-driven approaches to lower end costs will make these essential services more accessible.

Considerations for Data Analytics in Senior Living

Senior living organizations hoping to capitalize on data’s potential should keep a few things in mind. Interoperability is among the most important, as these businesses implement a wider range of electronic devices and services. Almost 90% of clinicians consult multiple electronic systems to access patient information, hindering efficiency, so LTC facilities should look for consolidated solutions providing a single access point.

Cybersecurity is another critical concern. There were over 700 major health care data breaches in 2022 alone, exposing millions of patient records. As LTC organizations increase their electronic data usage, they must adhere to strict access policies and implement advanced security safeguards to prevent these breaches.

Finally, senior living facilities must remember data-driven approaches only yield reliable results if the data itself is accurate. Investing in data verification and cleansing systems is a worthwhile endeavor to prevent losses from inaccurate or incomplete records.

Data Initiatives Can Boost Senior Standards of Living

When LTC organizations capitalize on their data, they can improve standards of living for their patients and make their companies more efficient. These advances benefit both the organizations themselves and their customers.

Data-driven approaches to senior care present a massive opportunity to the industry. As more LTC facilities become aware of and act on this potential, it will transform the sector for the better.

How to speed up claims processing with automated car damage detection

November 14, 2022/in Artificial Intelligence, Data Science, Deep Learning, Insights, Machine Learning, Main Category, Use Cases/by Benjamin Aunkofer

AI drives automation, not only in industrial production or for autonomous driving, but above all in dealing with bureaucracy. It is an realy enabler for lean management!

One example is the use of Deep Learning (as part of Artificial Intelligence) for image object detection. A car insurance company checks the amount of the damage by a damage report after car accidents. This process is actually performed by human professionals. With AI, we can partially automate this process using image data (photos of car damages). After an AI training with millions of photos in relation to real costs for repair or replacement, the cost estimation gets suprising accurate and supports the process in speed and quality.

AI drives automation and DATANOMIQ drives this automation with you! You can download the Infographic as PDF.

How to speed up claims processing
with automated car damage detection

Download this Infographic as PDF now by clicking here!

We wrote this article in cooperation with pixolution, a company for computer vision and AI-bases visual search. Interested in introducing AI / Deep Learning to your organization? Do not hesitate to get in touch with us!

DATANOMIQ is the independent consulting and service partner for business intelligence, process mining and data science. We are opening up the diverse possibilities offered by big data and artificial intelligence in all areas of the value chain. We rely on the best minds and the most comprehensive method and technology portfolio for the use of data for business optimization.

How Do Various Actor-Critic Based Deep Reinforcement Learning Algorithms Perform on Stock Trading?

May 27, 2022/in Artificial Intelligence, Deep Learning, Machine Learning, Main Category, Use Case, Use Cases/by Zhuotian Tang

Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy

Abstract

Deep Reinforcement Learning (DRL) is a blooming field famous for addressing a wide scope of complex decision-making tasks. This article would introduce and summarize the paper “Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy”, and discuss how these actor-critic based DRL learning algorithms, Proximal Policy Optimization (PPO), Advantage Actor Critic (A2C), and Deep Deterministic Policy Gradient (DDPG), act to accomplish automated stock trading by boosting investment return.

1 Motivation and Related Technology

It has long been challenging to design a comprehensive strategy for capital allocation optimization in a complex and dynamic stock market. With development of Artificial Intelligence, machine learning coupled with fundamentals analysis and alternative data has been in trend and provides better performance than conventional methodologies. Reinforcement Learning (RL) as a branch of it, is able to learn from interactions with environment, during which the agent continuously absorbs information, takes actions, and learns to improve its policy regarding rewards or losses obtained. On top of that, DRL utilizes neural networks as function approximators to approximate the Q-value (the expected reward of each action) in RL, which in return adjusts RL for large-scale data learning.

In DRL, the critic-only approach is capable for solving discrete action space problems, calculating Q-value to learn the optimal action-selection policy. On the other side, the actor-only approach, used in continuous action space environments, directly learns the optimal policy itself. Combining both, the actor-critic algorithm simultaneously updates the actor network representing the policy, and critic network representing the value function. The critic estimates the value function, while the actor updates the policy guided by the critic with policy gradients.

Figure 1: Overview of reinforcement learning-based stock theory.

2 Mathematical Modeling

2.1 Stock Trading Simulation

Given the stochastic nature of stock market, the trading process is modeled as a Markov Decision Process (MDP) as follows:

State s = [p, h, b]: a vector describing the current state of the portfolio consists of D stocks, includes stock prices vector p, the stock shares vector h, and the remaining balance b.
Action a: a vector of actions which are selling, buying, or holding (Fig.2), resulting in decreasing, increasing, and no change of shares h, respectively. The number of shares been transacted is recorded as k.
Reward r(s, a, s’): the reward of taking action a at state s and arriving at the new state s’.
Policy π(s): the trading strategy at state s, which is the probability distribution of actions.
Q-value : the expected reward of taking action a at state s following policy π.

A starting portfolio value with three actions result in three possible portfolios. Note that “hold” may lead to different portfolio values due to the changing stock prices.

Besides, several assumptions and constraints are proposed for practice:

Market liquidity: the orders are rapidly executed at close prices.
Nonnegative balance: the balance at time t+1 after taking actions at t, equals to the original balance plus the proceeds of selling minus the spendings of buying:
Transaction cost: assume the transaction costs to be 0.1% of the value of each trade:
Risk-aversion: to control the risk of stock market crash caused by major emergencies, the financial turbulence index that measures extreme asset price movements is introduced:

where denotes the stock returns, µ and Σ are respectively the average and covariance of historical returns. When exceeds a threshold, buying will be halted and the agent sells all shares. Trading will be resumed once returns to normal level.

2.2 Trading Goal: Return Maximation

The goal is to design a trading strategy that raises agent’s total cumulative compensation given by the reward function:

and then considering the transition of the shares and the balance defined as:

the reward can be further decomposed:

where:

At inception, h and $Q_{\pi}(s,a)$ are initialized to 0, while the policy π(s) is uniformly distributed among all actions. Afterwards, everything is updated through interacting with the stock market environment. By the Bellman Equation, $Q_{\pi}(s_t, a_t)$ is the expectation of the sum of direct reward $r(s_t,a_t,s_{t+1}$ and the future reqard $Q_{\pi}(s{t+1}, a_{a+1})$ at the next state discounted by a factor γ, resulting in the state-action value function:

2.3 Environment for Multiple Stocks

OpenAI gym is used to implement the multiple stocks trading environment and to train the agent.

State Space: a vector $[b_t, p_t, h_t, M_t, R_t, C_t, X_t]$ storing information about
$b_t$ : Portfolio balance
$p_t$ : Adjusted close prices
$h_t$ : Shares owned of each stock
$M_t$ : Moving Average Convergence Divergence
$R_t$ : Relative Strength Index
$C_t$ : Commodity Channel Index
$X_t$ : Average Directional Index
Action Space: {−k, …, −1, 0, 1, …, k} for a single stock, whose elements representing the number of shares to buy or sell. The action space is then normalized to [−1, 1], since A2C and PPO are defined directly on a Gaussian distribution.

Overview of the load-on-demand technique.

Furthermore, a load-on-demand technique is applied for efficient use of memory as shown above.

Algorithms Selection

This paper mainly uses the following three actor-critic algorithms:

A2C: uses parallel copies of the same agent to update gradients for different data samples, and a coordinator to pass the average gradients over all agents to a global network, which can update the actor and the critic network, with the objective function:
where $\pi_{\theta}(a_t|s_t)$ is the policy network, and $A(S_t|a_t)$ is the advantage function to reduce the high variance of it:
$V(S_t)$ is the value function of state $S_t$ , regardless of actions. DDPG: combines the frameworks of Q-learning and policy gradients and uses neural networks as function approximators; it learns directly from the observations through policy gradient and deterministically map states to actions. The Q-value is updated by:
Critic network is then updated by minimizing the loss function:
PPO: controls the policy gradient update to ensure that the new policy does not differ too much from the previous policy, with the estimated advantage function and a probability ratio:

The clipped surrogate objective function:

takes the minimum of the clipped and normal objective to restrict the policy update at each step and improve the stability of the policy.

An ensemble strategy is finally proposed to combine the three agents together to build a robust trading strategy. After training and testing the three agents concurrently, in the trading stage, the agent with the highest Sharpe ratio in one period will be automatically selected to use in the next period.

Implementation: Training and Validation

The historical daily trading data comes from the 30 DJIA constituent stocks.

Stock data splitting in-sample and out-of-sample.

In-sample training stage: data from 01/01/2009 – 09/30/2015 used to train 3 agents using PPO, A2C, and DDPG;
In-sample validation stage: data from 10/01/2015 – 12/31/2015 used to validate the 3 agents by 5 metrics: cumulative return, annualized return, annualized volatility, Sharpe ratio, and max drawdown; tune key parameters like learning rate and number of episodes;
Out-of-sample trading stage: unseen data from 01/01/2016 – 05/08/2020 to evaluate the profitability of algorithms while continuing training. In each quarter, the agent with the highest Sharpe ratio is selected to act in the next quarter, as shown below.
Table 1 – Sharpe Ratios over time.

Results Analysis and Conclusion

From Table II and Fig.5, one can notice that PPO agent is good at following trend and performs well in chasing for returns, with the highest cumulative return 83.0% and annual return 15.0% among the three agents, indicating its appropriateness in a bullish market. A2C agent is more adaptive to handle risk, with the lowest annual volatility 10.4% and max drawdown −10.2%, suggesting its capability in a bearish market. DDPG generates the lowest return among the three, but works fine under risk, with lower annual volatility and max drawdown than PPO. Apparently all three agents outperform the two benchmarks.

Table 2 – Performance Evaluation Comparison.

Cumulative return curves of our ensemble strategy and three actor-critic based algorithms, the min-vaiance portfolio allocation strategy, and the Dow Jones Industrial Average. (Initial portfolio value <img loading=

1,000,000, from 2016/01/04 to 2020/05/08).

Moreover, it is obvious in Fig.6 that the ensemble strategy and the three agents act well during the 2020 stock market crash, when the agents successfully stops trading, thus cutting losses.

Performance during the stock market crash in the first quarter of 2020.

From the results, the ensemble strategy demonstrates satisfactory returns and lowest volatilities. Although its cumulative returns are lower than PPO, it has achieved the highest Sharpe ratio 1.30 among all strategies. It is reasonable that the ensemble strategy indeed performs better than the individual algorithms and baselines, since it works in a way each elemental algorithm is supplementary to others while balancing risk and return.

For further improvement, it will be inspiring to explore more models such as Asynchronous Advantage Actor-Critic (A3C) or Twin Delayed DDPG (TD3), and to take more fundamental analysis indicators or ESG factors into consideration. While more sophisticated models and larger datasets are adopted, improvement of efficiency may also be a challenge.

Automated product quality monitoring using artificial intelligence deep learning

How to maintain product quality with deep learning

May 5, 2022/in Artificial Intelligence, Data Science, Deep Learning, Machine Learning, Main Category, Use Case, Use Cases/by Benjamin Aunkofer

Deep Learning helps companies to automate operative processes in many areas. Industrial companies in particular also benefit from product quality assurance by automated failure and defect detection. Computer Vision enables automation to identify scratches and cracks on product item surfaces. You will find more information about how this works in the following infografic from DATANOMIQ and pixolution you can download using the link below.

How to maintain product quality with automatic defect detection – Infographic

Download Infographic as PDF

Variational Autoencoders

April 19, 2022/in Artificial Intelligence, Data Science, Deep Learning, Machine Learning, Main Category, Use Cases/by Sunil Yadav

After Deep Autoregressive Models and Deep Generative Modelling, we will continue our discussion with Variational AutoEncoders (VAEs) after covering up DGM basics and AGMs. Variational autoencoders (VAEs) are a deep learning method to produce synthetic data (images, texts) by learning the latent representations of the training data. AGMs are sequential models and generate data based on previous data points by defining tractable conditionals. On the other hand, VAEs are using latent variable models to infer hidden structure in the underlying data by using the following intractable distribution function:

(1) $\begin{equation*} p_\theta(x) = \int p_\theta(x|z)p_\theta(z) dz. \end{equation*}$

The generative process using the above equation can be expressed in the form of a directed graph as shown in Figure ?? (the decoder part), where latent variable $z\sim p_\theta(z)$ produces meaningful information of $x \sim p_\theta(x|z)$ .

Figure 1: Architectures AE and VAE based on the bottleneck architecture. The decoder part work as
a generative model during inference.

Autoencoders

Autoencoders (AEs) are the key part of VAEs and are an unsupervised representation learning technique and consist of two main parts, the encoder and the decoder (see Figure ??). The encoders are deep neural networks (mostly convolutional neural networks with imaging data) to learn a lower-dimensional feature representation from training data. The learned latent feature representation $z$ usually has a much lower dimension than input $x$ and has the most dominant features of $x$ . The encoders are learning features by performing the convolution at different levels and compression is happening via max-pooling.

On the other hand, the decoders, which are also a deep convolutional neural network are reversing the encoder’s operation. They try to reconstruct the original data $x$ from the latent representation $z$ using the up-sampling convolutions. The decoders are pretty similar to VAEs generative models as shown in Figure 1, where synthetic images will be generated using the latent variable $z$ .

During the training of autoencoders, we would like to utilize the unlabeled data and try to minimize the following quadratic loss function:

(2) $\begin{equation*} \mathcal{L}(\theta, \phi) = ||x-\hat{x}||^2, \end{equation*}$

The above equation tries to minimize the distance between the original input and reconstructed image as shown in Figure 1.

Variational autoencoders

VAEs are motivated by the decoder part of AEs which can generate the data from latent representation and they are a probabilistic version of AEs which allows us to generate synthetic data with different attributes. VAE can be seen as the decoder part of AE, which learns the set parameters $\theta$ to approximate the conditional $p_\theta(x|z)$ to generate images based on a sample from a true prior, $z\sim p_\theta(z)$ . The true prior $p_\theta(z)$ are generally of Gaussian distribution.

Network Architecture

VAE has a quite similar architecture to AE except for the bottleneck part as shown in Figure 2. in AES, the encoder converts high dimensional input data to low dimensional latent representation in a vector form. On the other hand, VAE’s encoder learns the mean vector and standard deviation diagonal matrix such that $z\sim \matcal{N}(\mu_z, \Sigma_x)$ as it will be performing probabilistic generation of data. Therefore the encoder and decoder should be probabilistic.

Training

Similar to AGMs training, we would like to maximize the likelihood of the training data. The likelihood of the data for VAEs are mentioned in Equation 1 and the first term $p_\theta(x|z)$ will be approximated by neural network and the second term $p(x)$ prior distribution, which is a Gaussian function, therefore, both of them are tractable. However, the integration won’t be tractable because of the high dimensionality of data.

To solve this problem of intractability, the encoder part of AE was utilized to learn the set of parameters $\phi$ to approximate the conditional $q_\phi (z|x)$ . Furthermore, the conditional $q_\phi (z|x)$ will approximate the posterior $p_\theta (z|x)$ , which is intractable. This additional encoder part will help to derive a lower bound on the data likelihood that will make the likelihood function tractable. In the following we will derive the lower bound of the likelihood function:

(3) $\begin{equation*} \begin{flalign} \begin{aligned} log \: p_\theta (x) = & \mathbf{E}_{z\sim q_\phi(z|x)} \Bigg[log \: \frac{p_\theta (x|z) p_\theta (z)}{p_\theta (z|x)} \: \frac{q_\phi(z|x)}{q_\phi(z|x)}\Bigg] \\ = & \mathbf{E}_{z\sim q_\phi(z|x)} \Bigg[log \: p_\theta (x|z)\Bigg] - \mathbf{E}_{z\sim q_\phi(z|x)} \Bigg[log \: \frac{q_\phi (z|x)} {p_\theta (z)}\Bigg] + \mathbf{E}_{z\sim q_\phi(z|x)} \Bigg[log \: \frac{q_\phi (z|x)}{p_\theta (z|x)}\Bigg] \\ = & \mathbf{E}_{z\sim q_\phi(z|x)} \Big[log \: p_\theta (x|z)\Big] - \mathbf{D}_{KL}(q_\phi (z|x), p_\theta (z)) + \mathbf{D}_{KL}(q_\phi (z|x), p_\theta (z|x)). \end{aligned} \end{flalign} \end{equation*}$

In the above equation, the first line computes the likelihood using the logarithmic of $p_\theta (x)$ and then it is expanded using Bayes theorem with additional constant $q_\phi(z|x)$ multiplication. In the next line, it is expanded using the logarithmic rule and then rearranged. Furthermore, the last two terms in the second line are the definition of KL divergence and the third line is expressed in the same.

In the last line, the first term is representing the reconstruction loss and it will be approximated by the decoder network. This term can be estimated by the reparametrization trick \cite{}. The second term is KL divergence between prior distribution $p_\theta(z)$ and the encoder function $q_\phi (z|x)$ , both of these functions are following the Gaussian distribution and has the closed-form solution and are tractable. The last term is intractable due to $p_\theta (z|x)$ . However, KL divergence computes the distance between two probability densities and it is always positive. By using this property, the above equation can be approximated as:

(4) $\begin{equation*} log \: p_\theta (x)\geq \mathcal{L}(x, \phi, \theta) , \: \text{where} \: \mathcal{L}(x, \phi, \theta) = \mathbf{E}_{z\sim q_\phi(z|x)} \Big[log \: p_\theta (x|z)\Big] - \mathbf{D}_{KL}(q_\phi (z|x), p_\theta (z)). \end{equation*}$

In the above equation, the term $\mathcal{L}(x, \phi, \theta)$ is presenting the tractable lower bound for the optimization and is also termed as ELBO (Evidence Lower Bound Optimization). During the training process, we maximize ELBO using the following equation:

(5) $\begin{equation*} \operatorname*{argmax}_{\phi, \theta} \sum_{x\in X} \mathcal{L}(x, \phi, \theta). \end{equation*}$

Furthermore, the reconstruction loss term can be written using Equation 2 as the decoder output is assumed to be following Gaussian distribution. Therefore, this term can be easily transformed to mean squared error (MSE).

During the implementation, the architecture part is straightforward and can be found here. The user has to define the size of latent space, which will be vital in the reconstruction process. Furthermore, the loss function can be minimized using ADAM optimizer with a fixed batch size and a fixed number of epochs.

Figure 2: The results obtained from vanilla VAE (left) and a recent VAE-based generative
model NVAE (right)

In the above, we are showing the quality improvement since VAE was introduced by Kingma and
Welling [KW14]. NVAE is a relatively new method using a deep hierarchical VAE [VK21].

Summary

In this blog, we discussed variational autoencoders along with the basics of autoencoders. We covered
the main difference between AEs and VAEs along with the derivation of lower bound in VAEs. We
have shown using two different VAE based methods that VAE is still active research because in general,
it produces a blurry outcome.

References

[KW14] Diederik P Kingma and Max Welling. Auto-encoding variational bayes, 2014.
[VK21] Arash Vahdat and Jan Kautz. Nvae: A deep hierarchical variational autoencoder, 2021.

Air Quality Forecasting Python Project

March 20, 2022/in Data Mining, Data Science, Python, Use Cases/by Shivani Padaya

You will find the full python code and all visuals for this article here in this gitlab repository. The repository contains a series of analysis, transforms and forecasting models frequently used when dealing with time series. The aim of this repository is to showcase how to model time series from the scratch, for this we are using a real usecase dataset

This project forecast the Carbon Dioxide (Co2) emission levels yearly. Most of the organizations have to follow government norms with respect to Co2 emissions and they have to pay charges accordingly, so this project will forecast the Co2 levels so that organizations can follow the norms and pay in advance based on the forecasted values. In any data science project the main component is data, for this project the data was provided by the company, from here time series concept comes into the picture. The dataset for this project contains 215 entries and two components which are Year and Co2 emissions which is univariate time series as there is only one dependent variable Co2 which depends on time. from year 1800 to year 2014 Co2 levels were present in the dataset.

The dataset used: The dataset contains yearly Co2 emmisions levels. data from 1800 to 2014 sampled every 1 year. The dataset is non stationary so we have to use differenced time series for forecasting.

After getting data the next step is to analyze the time series data. This process is done by using Python. The data was present in excel file so first we need to read that excel file. This task is done by using Pandas which is python libraries to creates Pandas Data Frame. After that preprocessing like changing data types of time from object to DateTime performed for the coding purpose. Time series contain 4 main components Level, Trend, Seasonality and Noise. To study this component, we need to decompose our time series so that we can batter understand our time series and we can choose the forecasting model accordingly because each component behave different on the model. also by decomposing we can identify that the time series is multiplicative or additive.

CO2 emissions – plotted via python pandas / matplotlib

Decomposing time series using python statesmodels libraries we get to know trend, seasonality and residual component separately. the components multiply together to make the time series multiplicative and in additive time series components added together. Taking the deep dive to understand the trend component, moving average of 10 steps were applied which shows nonlinear upward trend, fit the linear regression model to check the trend which shows upward trend. talking about seasonality there were combination of multiple patterns over time period which is common in real world time series data. capturing the white noise is difficult in this type of data. the time series contains values from 1800 where the Co2 values are less then 1 because of no human activities so levels were decreasing. By the time numbers of industries and human activities are rapidly increasing which causes Co2 levels rapidly increasing. In time series the highest Co2 emission level was 18.7 in 1979. It was challenging to decide whether to consider this values which are less then 0.5 as white noise or not because 30% of the Co2 values were less then 1, in real world looking at current scenario the chances of Co2 emission level being 0 is near to impossible still there are chances that Co2 levels can be 0.0005. So considering each data point as a valuable information we refused to remove that entries.

Next step is to create Lag plot so we can see the correlation between the current year Co2 level and previous year Co2 level. the plot was linear which shows high correlation so we can say that the current Co2 levels and previous levels have strong relationship. the randomness of the data were measured by plotting autocorrelation graph. the autocorrelation graph shows smooth curves which indicates the time series is non–stationary thus next step is to make time series stationary. in non–stationary time series, summary statistics like mean and variance change over time.

To make time series stationary we have to remove trend and seasonality from it. Before that we use dickey fuller test to make sure our time series is non–stationary. the test was done by using python, and the test gives p–value as output. here the null hypothesis is that the data is non–stationary while alternate hypothesis is that the data is stationary, in this case the significance values is 0.05 and the p–values which is given by dickey fuller test is greater than 0.05 hence we failed to reject null hypothesis so we can say the time series is non–stationery. Differencing is one of the techniques to make time series stationary. On this time series, first order differencing technique applied to make the time series stationary. In first order differencing we have to subtract previous value from current value for all the data points. also different transformations like log, sqrt and reciprocal were applied in the context of making the time series stationary. Smoothing techniques like simple moving average, exponential weighted moving average, simple exponential smoothing and double exponential smoothing techniques can be applied to remove the variation between time stamps and to see the smooth curves.

Smoothing techniques also used to observe trend in time series as well as to predict the future values. But performance of other models was good compared to smoothing techniques. First 200 entries taken to train the model and remaining last for testing the performance of the model. performance of different models measured by Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) as we are predicting future Co2 emissions so basically it is regression problem. RMSE is calculated by root of the average of squared difference between actual values and predicted values by the model on testing data. Here RMSE values were calculated using python sklearn library. For model building two approaches are there, one is data–driven and another one is model based. models from both the approaches were applied to find the best fitted model. ARIMA model gives the best results for this kind of dataset as the model were trained on differenced time series. The ARIMA model predicts a given time series based on its own past values. It can be used for any non–seasonal series of numbers that exhibits patterns and is not a series of random events. ARIMA takes 3 parameters which are AR, MA and the order of difference. Hyper parameter tuning technique gives best parameters for the model by trying different sets of parameters. Although The autocorrelation and partial autocorrelation plots can be use to decide AR and MA parameter because partial autocorrelation function shows the partial correlation of a stationary time series with its own lagged values so using PACF we can decide the value of AR and from ACF we can decide the value of MA parameter as ACF shows how data points in a time series are related.

Yearly difference of CO2 emissions – ARIMA Prediction

Apart from ARIMA, few other model were trained which are AR, ARMA, Simple Linear Regression, Quadratic method, Holts winter exponential smoothing, Ridge and Lasso Regression, LGBM and XGboost methods, Recurrent neural network (RNN) – Long Short Term Memory (LSTM) and Fbprophet. I would like to mention my experience with LSTM here because it is another model which gives good result as ARIMA. the reason for not choosing LSTM as final model is its complexity. As ARIMA is giving appropriate results and it is simple to understand and requires less dependencies. while using lstm, lot of data preprocessing and other dependencies required, the dataset was small thus we used to train the model on CPU, otherwise gpu is required to train the LSTM model. we face one more challenge in deployment part. the challenge is to get the data into original form because the model was trained on differenced time series, so it will predict the future values in differenced format. After lot of research on the internet and by deeply understanding mathematical concepts finally we got the solution for it. solution for this issue is we have to add previous value from the original data from into first order differencing and then we have to add the last value of this time series into predicted values. To create the user interface streamlit was used, it is commonly used python library. the pickle file of the ARIMA model were used to predict the future values based on user input. The limit for forecasting is the year 2050. The project was uploaded on google cloud platform. so the flow is, first the starting year from which user want to forecast was taken and the end year till which year user want to forecast was taken and then according to the range of this inputs the prediction takes place. so by taking the inputs the pickle file will produce the future Co2 emissions in differenced format, then the values will be converted to original format and then the original values will be displayed on the user interface as well as the interactive line graph were displayed on the interface.

You will find the full python code and all visuals for this article here in this gitlab repository.

How to ensure occupational safety using Deep Learning – Infographic

March 3, 2022/in Artificial Intelligence, Data Science, Deep Learning, Insights, Machine Learning, Main Category, Use Cases/by Benjamin Aunkofer

In cooperation between DATANOMIQ, my consulting company for data science, business intelligence and process mining, and Pixolution, a specialist for computer vision with deep learning, we have created an infographic (PDF) about a very special use case for companies with deep learning: How to ensure occupational safety through automatic risk detection using using Deep Learning AI.

How to ensure occupational safety through automatic risk detection using Deep Learning – Infographic

Infographic as PDF Download