Introduction to Analyzing SIP Inflows in India
To understand the key drivers of SIP (Systematic Investment Plan) inflows in India, three distinct analytical approaches are adopted: The Feature Selection Approach, The Ordinary Least Squares (OLS) Regression Approach, and The SHapley Additive exPlanations (SHAP) Approach. The primary objective here is not forecasting but identifying the most influential feature driving SIP inflows, making time-wise sequencing of data less of a focus.
The Feature Selection Approach
Feature selection is a crucial step in machine learning that helps extract the most important features when dealing with numerous variables. There are three primary methods under feature selection: filter, wrapper, and embedded methods.
Filter Methods
Filter methods select features based on their scores in statistical tests for their correlation with the target variable, SIP inflows. Given the continuous nature of the target and feature variables, Pearson’s Correlation is applied. However, before applying Pearson’s Correlation, it’s essential to consider multicollinearity among features.
Understanding Multicollinearity and Correlation
A correlation matrix revealed a nearly perfect positive correlation between CPI (Combined) and the NIFTY50 Index. This relationship seems counterintuitive because rising inflation (CPI) would typically lead to higher interest rates, negatively impacting stock prices. However, this positive relationship might reflect a period where inflation was accompanied by economic growth. To resolve this, Pearson’s correlation is used to determine which of the two has a higher correlation with SIP inflows. The NIFTY50 Index has the highest correlation with SIP inflows, at 0.95, making it the key feature influencing SIP inflows according to the filter method.
Wrapper Methods
Wrapper methods evaluate combinations of features by testing which ones perform best in a machine learning model. Using the Random Forest Regressor, CPI (Combined) emerges with a relative importance of 0.74, indicating its dominant role in predicting SIP inflows. Theoretically, this makes sense as higher inflation rates would encourage individuals to invest in higher-risk instruments for better returns. Thus, the wrapper method identifies CPI (Combined) as the key feature.
Embedded Methods
Embedded methods integrate feature selection into the model-building process. Methods include Lasso, Ridge, and ElasticNet Regression. All three methods identified the NIFTY50 Index as the most important feature, with the repo rate being of the least importance. This confirms the NIFTY50 Index as the key feature influencing SIP inflows according to the embedded methods.
Conclusion
The feature selection approach, through its filter, wrapper, and embedded methods, provides insight into the key variable driving SIP inflows in India. While the wrapper method suggests CPI (Combined) as influential, both filter and embedded methods consistently highlight the NIFTY50 Index as the primary driver of SIP inflows. This indicates that the performance of the NIFTY50 Index has a significant impact on individual investment decisions regarding SIPs, as a rising index tends to stimulate market investment. Understanding these drivers can help in predicting and managing SIP inflows more effectively, contributing to a more stable and growing investment environment in India.
