Explainable Inflation Forecasts by Machine Learning Models

Abstract

A comprehensive analysis of inflation prediction using a variety of machine learning techniques, focusing on identifying the key factors for accurate forecasting. By analyzing data from the World Bank and Kaggle databases from 1970 to 2022, we assessed the predictive power of different aspects of inflation such as Food Inflation, Energy Inflation, Headline Consumer Inflation, Producer Inflation, and Official Core Consumer Inflation. Our methodology included extensive data cleaning, feature engineering, and model training to predict inflation trends over the next 10-12 years using models like Linear Regression, Random Forest, Decision Tree, AdaBoost, XGBoost, Ridge, Lasso, and Gradient Boosting. We also employed techniques like hyperparameter tuning, performance evaluation metrics, and result visualization to gain deeper insights into our research findings. The findings indicate that Producer Price Inflation, Food Inflation, and Energy Inflation play a crucial role in predicting future trends. Rising rates in these sectors indicate broader economic shifts, underscoring the necessity for targeted interventions. This research underscores the importance of utilizing a range of inflation indicators to enhance the accuracy and robustness of forecasting models. Our study significantly enhances economic forecasting by showcasing the efficacy of explainable machine learning models in predicting inflation trends and underscores the importance of grasping inflation dynamics. This promotes a deeper understanding and well-informed decision-making among policymakers, businesses, and the public, fostering proactive measures to stabilize the economy and mitigate the adverse effects of inflation.