Forecasting the Number of Road Accidents in Major and Secondary Tourist Cities in Thailand

Abstract

This study purpose to identify the most suitable forecasting model and compare the forecasting performance of different methods to determine the best approach for predicting the number of road accidents in Thailand. Additionally, this study considers the impact of tourism promotion and regional income distribution by categorizing data into two groups, following the classification of the Tourism Authority of Thailand (TAT): (1) Major Tourist Cities and (2) Secondary Tourist Cities. The forecasting models were developed using monthly road accident data collected from the Ministry of Transport’s accident reporting system, managed by the Information and Communication Technology Center, Office of the Permanent Secretary, Ministry of Transport. The dataset spans from January 2019 to December 2024, Totaling 72 months. This study employs 3 individual forecasting methods based on time series analysis: (1) Exponential Smoothing Method, (2) Time Series Regression Method, and (3) Box-Jenkins Method. Additionally,
5 combined forecasting methods: (1) Simple Average Method or Equivalent Weighted method: EW,
(2) Regression-Based combination, (3) MAPE-Based combination, (4) Unequivalent Weighted Method: UNEW, and (5) Weighted Ranking method: WR. The forecasting accuracy of each method was assessed using the Mean Absolute Percentage Error (MAPE) criterion, with the model yielding the lowest MAPE being selected as the most effective.

       The findings indicate that for Major tourist cities, the most efficient forecasting approach was the Unequivalent Weighted Method (UNEW). While, Secondary tourist cities, the best-performing method was the Regression-Based Combination. The results of this study are expected to be beneficial for relevant authorities in formulating effective road accident prevention measures and response strategies to enhance public safety and regional planning.