Statistical Methods and Machine Learning Approaches for Predicting Basic Commodity Prices in East Java

Abstract

Price fluctuations of basic commodities impact economic stability and community welfare. This study compares predictive methods based on statistical approaches (Simple Moving Average, Linear Regression) and machine learning techniques (Support Vector Regression, Long Short-Term Memory) using data from SISKAPERBAPO, which records daily prices of 76 basic commodities across 119 central markets in 38 districts/cities in East Java. The study supports the role of Regional Inflation Control Teams (TPID) in maintaining stable and low inflation through coordinated policies. Evaluation based on Root Mean Square Error (RMSE) and Squared Correlation indicates that SVR performs best of 4 commodities (rice, sugar, chicken meat, chicken eggs), while LSTM excels for 3 commodities (cooking oil, beef, garlic). These findings recommend SVR and LSTM as the most efffective methods for price prediction and provide a reference for TPID and policymakers in developing for price control.