Ratio analysis is a cornerstone of financial statement interpretation and performance evaluation in banking institutions. It provides a structured framework to examine profitability, liquidity, solvency, and operational efficiency. In this project, we perform a detailed ratio analysis of ICICI Bank, one of India’s leading private-sector banks, and extend this analysis using modern Machine Learning (ML) and Deep Learning (DL) techniques. The aim is to transform ratio analysis from a static, backward-looking exercise to a dynamic, predictive financial intelligence tool. The study employs traditional financial ratios such as Net Profit Margin, Return on Assets (ROA), Capital Adequacy Ratio (CAR), Gross and Net NPA Ratios, Credit-Deposit Ratio (CDR), and Cost-Income Ratio. These ratios are examined over a 10-year span using data sourced from ICICI Bank’s annual reports, RBI disclosures, and financial APIs. Patterns and anomalies in the ratio behavior are interpreted using Exploratory Data Analysis (EDA).To introduce predictive capabilities, ML models like Random Forest, Support Vector Regression (SVR), and XGBoost are trained on historical ratio data to forecast key metrics like ROA and NPAs. Furthermore, time-series DL models like Long Short-Term Memory (LSTM) and GRU are applied to forecast financial ratios quarterly, capturing complex patterns that traditional methods miss. A Streamlit-based dashboard was also developed to enable real-time, user-friendly interaction with the data and prediction outputs. The project concludes that AIenhanced ratio analysis significantly improves the depth, accuracy, and foresight of traditional techniques, offering a strategic advantage to financial analysts, investors,and regulators alike.
Kumar et al. (Wed,) studied this question.