# Data Science: The Art and Science of turning Complex Calculations into Intuitive and Actionable Analytics

Nathalie Leroy
August 21, 2024

The integration of data science, statistical methods, mathematical algorithms, and probabilistic models into real-time operational analytics can transform how organizations function. SaaS solutions are able to integrate advanced analytics, providing users with continuous monitoring, predictive insights, and efficient decision-making processes.‍ These techniques allow non-technical users to access results in an intuitive and understandable format based on sophisticated mathematical methods. They can therefore make decisions and perform actions within the context of the solution based on robust underlying analysis.

Analytics
Data Visualization
Artificial Intelligence
Dashboards
KPI

Organizations are increasingly leveraging advanced data science techniques to enhance their operational efficiencies and make informed decisions in real time.

Imagine an industry-specific solution (logistics, finance, healthcare, retail, CRM, etc) that aims to integrate advanced analytics. These analytics are provided through dashboards, offering valuable insights directly within user workflows.

But what do these analytics actually consist of? They can range from basic calculations to sophisticated mathematical methods.

This blog post explores how data science, statistical methods, mathematical algorithms, and probabilistic models can be incorporated into these industry-specific solutions in an easy way for non-technical users to consume.

We will first give a quick background on these mathematical techniques, and then illustrate their practical applications through two examples: a logistics company and a healthcare solution provider.

## Mathematical Background

Data science is rooted in various mathematical disciplines, here are a few key concepts:

• Statistical Methods: Techniques such as regression analysis, hypothesis testing, and time series analysis are essential for identifying trends, making predictions, and inferring relationships from data.
• Mathematical Algorithms: Algorithms like k-means clustering and decision trees are used to classify, cluster, and predict outcomes based on historical data.
• Probabilistic Models: Models such as Binomial / Poisson distributions, Bayesian inference and Markov chains handle uncertainty and make predictions about future events based on past data.

In the industry-specific software analytics context, the above can be applied “in production”, meaning having dashboards show results of those techniques in a readable, understandable and intuitive format for any type of user. These techniques will turn into Real-Time Analytics, meaning algorithms and models will continuously be applied to new data streams. Users will gain immediate insights based on a powerful analytical basis.

## Diving into two stories of icCube Customers using those Techniques.

### IoT in Logistics: Enhancing Fleet Management with Integrated Analytics

A UK-based company leasing trains to national railroads, needed to maintain the quality and reliability of their fleet due to regulatory policies. To address this, they sought to integrate Internet of Things (IoT) technology with a comprehensive analytics solution for real-time monitoring and analysis.

icCube developed a custom data source for accessing data from their Metron temperature sensors installed in the train wagons. These sensors continuously measure the internal temperatures and other KPIs, providing a constant stream of data read by their systems.

Together, we created several dashboards to supervise the performance of the heating and cooling systems. These dashboards compare the inside temperature of the wagons with the outside temperature to detect anomalies, indicating specific trains that require attention. The dashboards are automatically refreshed with new data, enabling quick decision-making.

###### Value

The underlying mathematical techniques applied, including correlation analysis and probabilistic models like the Binomial distribution and decision trees, allows the different users of the solution to benefit from them:

• On the one hand, train operators can detect anomalies in the temperature regulation system in real time and immediately act on the malfunctioning trains.
• On the other hand, top management is provided with historical patterns and predictive results in order to schedule maintenance procedures and prevent issues from occurring. Moreover, automatic weekly PDF reports with the results mentioned above are generated by icCube and sent to top management, detailing the performance of the temperature regulation systems.

By leveraging advanced mathematical models embedded within the solution, the logistics company is able to enable real-time anomaly detection and predictive maintenance.

### Healthcare Solution: Prioritizing Patient Care with Advanced Analytics

A healthcare solution provider sought to leverage data science and advanced analytics integrated within their solution to meet regulatory standards and prioritize patient treatment effectively.

The company needed an analytics solution to enable real-time prioritization of patients requiring urgent care. The solution embedded icCube as its analytical and data visualization layer, enabling users to perform advanced slice & dice techniques in order to cluster patients, and take immediate actions, e.g. bulk schedule appointments for thousands of patients in seconds.

Another use case within this solution is the usage of behavior analysis, predicting outcomes based on patient history and behavior. The usage of advanced mathematical models such as linear regressions, allow Management to define the patients’ programs investment strategies.

###### Value

Through better scoring and understanding of patients, better assignment of the right, prioritized patients to the right healthcare programs (for instance: chronic care, diabetes, mental illness and others or combinations thereof), can be made.

Consider, for instance, there might be 1,000 funded places available in a single program, but 10,000 patients that could benefit from the program.  Which of the 10,000 patients should be assigned to the program (today) ensuring all 1,000 funded places are actually used. There is perhaps no perfect answer, there are many, often time-bound, variables, which modeled, weighted, scored and calculated make the best possible use of finite resources availability.  In other words, putting “the brains” in the analytic engine.

There’s also the challenge of scheduling all this. To do this manually would be time-consuming, inefficient and expensive, so embedding these analytic capabilities into an application that can generate all the necessary parameterised notifications and assignments is key.

Since these programs are funded by the tax-payer grants, tax-payer subsidized sources or insurers, this ensures the healthcare provider is balancing best possible patient-outcomes along with maximum return on funding/grants. Furthermore, transparency is provided to ensure the healthcare provider is seen to fulfill its regulatory obligations and therefore avoiding violations subject to large financial penalties to the tune of millions of dollars as well as (detrimental to all concerned stakeholders) loss of future support.

Finally, these analytics can be put to work by the healthcare provider with funding sources as evidence justifying future programs and healthcare budgeting to support their successful programs whilst at the same time generating virtuous circles of continuous improvement for patients, funding sources, the government and ultimately the electorate and society as a whole.

Yes, we knew you were going to ask the question :) Data science and AI are interconnected but serve different purposes.

It’s a huge topic, so if we focus on the context above, there are different ways AI could be applied in those use cases. If we’re specifically thinking of AI as a “Data Scientist Helper”, nothing prevents it from achieving the same mathematical result as the ones above. But as of today, we would need to guide the AI to achieve the needed result or make (mathematical) model suggestions, as it’s not only about the math knowledge but also the expertise of the industry specialists. So it’s a combination of scientific knowledge, business knowledge, professional experiences, and last but not least, the industry-solution provider's choice. Read more about this in our Blog Post about LLM and Data Analytics.

The data science techniques explained above and applied to those two use cases were chosen methods those companies decided to apply to their data as a default analysis to be done.

It’s up to the solution provider to decide what type of analysis to provide to their users, whether it’s underlying analysis done by internal data scientists and experts and/or opening up an AI solution to users to analyze data. Topic to be continued :)

## Conclusion

The integration of data science, statistical methods, mathematical algorithms, and probabilistic models into real-time operational analytics can transform how organizations function. SaaS solutions are able to integrate advanced analytics, providing users with continuous monitoring, predictive insights, and efficient decision-making processes.

These techniques allow non-technical users to access results in an intuitive and understandable format based on sophisticated mathematical methods. They can therefore make decisions and perform actions within the context of the solution based on robust underlying analysis.