A branch of data science that is not often prioritized

With the rapid adoption of AI across all domains, there are only a handful of areas that the technology is not part of. Interestingly, the role of AI and data science in one particular area – operations research – is highly ambiguous. A common perception is that operations research (OR) is not useful for data science; moreover, the overlap between data science and OR is poorly understood.

This misconception stems mainly from the marketing of OR products and services that are applied to the real world – very often end users do not understand the terms OR and data science. Another reason is that readily available machine learning models are available as multi-platform packages like Python and don’t really contain specific OR models.

In reality, however, OR problems are applicable to AI and data science. In fact, many ideas in AI and data science problem solving have intersected from OR due to the large overlap in techniques and methods used.

Speaking of the same, Rajeev Rajan, AVP, Data Science, Genpact, at the MLDS 2022 event.

Operations research and machine learning

Operations research is a fundamental part of the overall machine learning lifecycle. It is especially useful when dealing with business issues that require parameter optimization. Some of the examples of OU are:

• Enable intelligent workforce management by forecasting resource requirements and optimizing the daily resource schedule
• Increase the audience of TV programs thanks to optimal programming of program promotion
• Enable supply chain transformation by providing AI/machine learning based recommendations for optimized product usage
• AI-powered forecasting for retail and e-commerce applications to optimize funnel and customer traffic
• Enable data-driven optimization for automated warehouse management, inspection and quality control

“We should think of artificial intelligence, data science and operations research as three different things. That said, even in machine learning there is a certain type of optimization, okay, a certain type of modeling math is in progress, and we need to estimate the parameter correctly. As an ML guide or data scientist, one should not think that a basic operations research problem is beyond their purview. If you can ignore this step, you might not get the full satisfaction that you were supposed to get from bringing that particular ML or particular AI solutions to customers,” says Rajan.

Rajan explained operations research and its role in the overall machine learning scheme through many examples. One such use case was for increasing the viewership of TV programs. The challenge here, as Rajan explained, is to apply machine learning techniques to avoid revenue loss. This revenue loss is seen as a result of a decline in viewership due to sub-optimal scheduling of promotions between programs.

To this end, Rajan offers a solution where one collects data and performs mixed integer programming in Python using the Gurobi solver to generate an optimal promotion schedule. This translates to a 4.5% increase in revenue due to increased viewership of promoted shows.

An AI development lifecycle includes the following stages:

• Kicking off AIDLC: This step consists of defining the problem to be solved.
• Idea and Assessment: This step involves understanding the current state and defining the scope of work accordingly.
• ML model development: Here the machine learning solution is developed and tested.
• ML outputs given as OR inputs: Here OR techniques are used to make recommendations based on the outputs of the ML model. This is a crucial step for the entire life cycle.
• Last step: Finally, the output of the solution is delivered to the customer.

Wrap

According to Rajan, here are some of the observations of integrating OR into an AI initiative:

• Data science and OR are not always seen as closely related. Most companies that leverage AI and advanced analytics employ multidisciplinary teams that cover both.
• Hybrid techniques of OR and data science are used effectively to deploy end-to-end solutions.
• When the end goal of customers is to automate decision-making, products are referred to as AI rather than OR.
• Using OR tools and techniques in AI applications will help spread AI integrations at scale.
• Developing OR as a skill set is an essential part of an effective AI initiative.