Abstract

The study critically reviews the existing literature on inflation forecasting in India, emphasizing the limitations of traditional econometric models in capturing intricate relationships among macroeconomic variables influencing inflation. In response, Machine Learning (ML) and Deep Learning (DL) models are proposed for inflation forecasting, leveraging a dataset comprising Indian macroeconomic variables. The devised model undergoes training and testing through various ML and DL techniques, encompassing LR, K-NN, DT, SVM (Linear Kernel), SVM (RBF Kernel), NN, RF, GB, AdaBoost, and XGBoost. Comparative analysis against traditional econometric models demonstrates the superior predictive performance of the ML and DL model. Notably, the convolutional neural network model stands out as the most accurate forecasting techniques. This study also identifies key variables influencing inflation rates in India, including interest rates, exchange rates, and fuel prices. The results underscore the potential of ML and DL methods in enhancing inflation forecasting in India, utilizing the latest dataset collected through survey methods across diverse Indian cities. Furthermore, XGBoost emerges as the optimal classifier for predicting inflation in India. These findings hold significant implications for policymakers and central banks, offering valuable insights for informed decision-making and the formulation of effective monetary policies in the context of developing countries.