Can I Combine Qualitative And Quantitative Methods?
February 20, 2020
“Will My Research Be Inductive Or Deductive?”
February 29, 2020Business Analytics & Operational Research: Decision Analysis, Linear Programming, Simulation & Markov Chains
- Home
- Blog
- Business Analytics & Operational Research: Decision Analysis, Linear Programming, Simulation & Markov Chains
Business Analytics & Operational Research: Decision Analysis, Linear Programming, Simulation & Markov Chains
In modern years, there is a growing demand in the field of business analytics. It actually means that what outcome we should get in business from the data to make better decisions. This is often sound like relating a business problem to an operation research problem. However, there is often a question arise in connecting the business analytics to the operation research problem. In this blog, I will explain you the meaning of business analytics and how it is related and useful in the operation research methods or decision making including linear programming, inventory management, simulation and Markov chains.
Analytics are used to identify (i) what has happened? (ii) What should happen? And (iii) what will happen? In the business. These three forms of question are categorized into Descriptive, Prescriptive and Predictive analytics respectively. However, business analytics is the study of data via statistical techniques, constructing predictive models, implementing the optimizing rule and draw a valid inference according to the business needs. Thus, business analytics uses a huge amount of data or simply big data to make a profitable conclusion.
There is a different approach to business analytics, which in turn
delivers profitable benefits (Budnick et al., 1994). I will list out a few uses of business analytics for the
betterment of the business.
- If a business company wants to identify the pattern of the sales of a product or to find a new pattern to promote the growth of the business, then business analytics is used to implement the data mining techniques such as classification, regression analysis, clustering analysis, etc., and to understand the complex data using neural networks, deep learning and machine learning techniques.
- Business analytics is used to do quantitative statistical analysis or solving a mathematical model to deliver justifications for the occurrence of the problem
- It can be used as a supporting tool for conducting any multivariate testing and A/B testing to find the relationship or test the relationship with past decisions.
- It can be used for predictive modelling to improve business standards.
Apart from the benefits and uses of business analytics, the main goal of business analytics is to identify which dataset will be useful and how it can be taken forward to solve the business problems and increase the profit, productivity, and efficiency. So far, I explained to you about the meaning and benefits of business analytics. However, in recent years, business analytics in operational practice has become a great interest among researchers.
With the growth of technologies, and with the large amount of data at hand, it is important to make use of analytics and the operation research approach to solve many complex business problems (Choi et al., 2017; Hillier & Lieberman, 2015). Thus, in the coming years, business analytics tools are the most powerful tool to take the business standard to the next level. Now, let look at how a simple Markov chain is used to solve a business problem.
Consider a bank which deals with both asset and liability products, and it is obvious that loans taken from the bank play a vital role in the revenue. Hence, the bank executive decided to hire a consultant to find whether they end up in good loans, risky loans, paid-up loans or bad loans.
In this example, the bad loans and the paid-up loans are the absorbing nodes or the end state in a Markov chain. The absorbing node is that it has no transition probability to any other nodes. So, as a statistical consultant, the first step is to understand the trends in the loan cycle with the previous study. Let’s say; the following Markov chain represents the pattern of loans for the previous year
From the above transition diagram, it is clear that the bad loans and paid-up loans are the absorbing states; that is, the process end and stays in these states forever. Otherwise, paid-up loans cannot be a bad loan or risky or good and similarly, the bad loans cannot be a paid-up or risky or good.
Next step is to calculate the transition probability matrix with the previous probability. That, it with the previous probability, estimate the number of loans belongs to each category. From the diagram, it is clear that 60% has good loans, and 40% has bad loans. Thus, the calculation becomes,
From the final output, it is expected that 15% of the loans are going to be paid-up loans for the current year and 16% becomes a bad loan. Thus, from this Markov chain example, the retail industry can develop their business insights to decrease the percentage of bad loans in the future. In addition, if you want to predict the same for 2 years, then with the same transition matrix, it is calculated as
Similarly, the process is repeated until the convergence is achieved. That is,
From the convergence result, it is identified that 54% of the present loan will be paid fully, and 46% will be a bad loan. This is useful in identifying the risk of banks in issuing loans to customers.
Suppose, if you want to identify the proportion of good loans becoming a paid loan, then you should start with 100% of good loans and others as 0% in the initial stage and repeat the process until convergence is achieved.
From the results, it is identified as only 23% becomes a bad loan whereas in the previous case it was recorded as 46%. Similarly, if you are interested in identifying the proportion of risk loans ending as paid-up or bad loans then assign 100% probability to risk loans and others with 0% probability the do the process until convergence and deliver a valid conclusion.
The previous case deals with the Markov process into business insights. However, there is still a question persists where the analytics relate to operation research? An operation research scientist is everywhere in the process and few having developed these kinds of tools to solve a business problem and few have developed a robust model for the same. In practice, Operational Analytics or business analytics involves building a suitable model or developing a predictive model to make meaningful business decisions. It may be a transportation model, or the Markov model, or the Linear programming model or a simulation model; the objective is to satisfy the business needs and do a profitable business.
Summary
I presented an informal description of business analytics and Operations Research in this blog with an application to a retail bank industry using Markov chains. I personally feel that if I want to understand anything, it is better to dig deeper into the topic and go for details.
References
- Budnick, F.S., McLeavey, D.W. & Mojena, R. (1994). Principles Of Operations Research For Management (2nd Edition). Irwin series in quantitative anlysis for business. [Online]. A.I.T.B.S. Publishers. Available from: https://books.google.co.in/books?id=wBMVYAAACAAJ.
- Choi, T.-M., Chan, H.K. & Yue, X. (2017). Recent Development in Big Data Analytics for Business Operations and Risk Management. IEEE Transactions on Cybernetics. [Online]. 47 (1). pp. 81–92. Available from: http://ieeexplore.ieee.org/
