PSU Bank Modeling - A Comparative Modeling Approach Involving Artificial Neural Network and Panel Data Regression

Indian Banking, especially the PSU segment has been facing the fire for quite some time now. Be it NPA, or be it digitization or even be it cash management, they are finding it hard to cope up with the changes. ‘Earnings per Share' (EPS) is a cardinal parameter which will impact the performance of the bank's share in the bourses. So, in this study the banks of PSU segment that are included in the defined universe of S&P BSE Bankex have been under consideration (namely SBI, BOB and PNB). Determination of EPS with high degree of accuracy will help these banks to attract FPI flow consistently which in turn will be beneficial for their regular cash inflow and thus ease their liquidity crunch. Two distinctintly different methods such as Panel Data Regression (Econometric Method) and Artificial Neural Network (Machine Learning Tool) have in in action to set up an accurate model, with which the EPS prediction could be quite accurate in nature. Thirteen control variables are chosen carefully to construct both the models. Interestingly both the models are depicting similar picture (with accuracy measures) with different combinations