Forecasting Indonesian food price inflation with multivariate time series and machine learning approaches

Food price inflation is a critical economic indicator significantly impacting public welfare. Accurate forecasting is essential, particularly in dynamic economies like Indonesia. This study forecasts Indonesian food price inflation using a multivariate time series approach combined with machine learning models. We utilized monthly data from the World Bank (Jan 2008-Dec 2023 for training/internal evaluation; Jan 2024-Dec 2024 for independent testing), including food price inflation and market prices (Open, High, Low, Close). We implemented and compared Support Vector Regression (SVR), Long Short-Term Memory (LSTM), and Bidirectional LSTM (BILSTM) models based on their Root Mean Square Error (RMSE). On the internal evaluation dataset, the SVR model demonstrated superior performance with an RMSE of 4.7990 compared to LSTM and BILSTM. When forecasting actual future data for 2024, the SVR model yielded an RMSE of 8.9238, indicating its capability in predicting unseen data. This study concludes that SVR is an effective model for forecasting Indonesian food price inflation, especially in scenarios with limited data and moderate complexity, providing valuable insights for strategic decision-making aimed at supporting economic stability and public welfare.