A Metaheuristic-Based Approach to Inflation Prediction in Indonesia with Support Vector Regression (SVR)

Abstract

Inflation prediction plays a vital role in individual, corporate, and government economic decision-making. This study evaluates the performance of the Support Vector Regression (SVR) algorithm in predicting monthly inflation using 251 data points from Bank Indonesia. Two parameter optimization methods, Grid Search Optimization and Particle Swarm Optimization (PSO), are applied to enhance prediction accuracy. Data pre-processing includes normalization using Min-Max Scaler and splitting into training and testing sets with ratios of 85:15 and 90:10. The optimization results indicate that PSO with a 90:10 ratio outperforms other approaches, achieving a Mean Absolute Percentage Error (MAPE) of 20.27% and an R-squared value of 89.76%. These findings highlight the significance of effective parameter optimization methods in improving prediction model performance, especially for non-linear data such as inflation. The results also demonstrate machine learning techniques' potential in analyzing time series data for economic and financial applications. This study provides insights into developing more accurate prediction systems, which can contribute to better economic planning and policy making.