Temperature Forecasting in Phnom Penh Using Time Series Models

Abstract

The weather or temperature data based on location is essential for business, society, agriculture, and energy applications. The main aim of this paper is to overview the data mining process for temperature data and to study temperature data using data mining techniques like time series forecasting that we can acquire. Four years of daily temperature data (2019- 2022) was collected based on stations. The temperature data, Phnom Penh, can be used to find the hidden patterns inside the large datasets to transfer the retrieved information into usable knowledge for predicting temperature conditions. Auto-Regressive Integrated Moving Average (ARIMA) and Seasonal Auto-Regressive Integrated Moving Average with eXogenous factors (SARIMAX) are time series data mining techniques generally used for time series analysis and future forecasting. Based on the results of the time series analysis, the Mean Absolute Percentage Error (MAPE) for the temperature data using ARIMA (1,0,3) was 0.006 and the Mean Absolute Percentage Error (MAPE) estimated for forecasting using SARIMAX (1,0,3) was 0.007. Thus, forecasting temperatures using ARIMA is slightly better performance than SARIMAX.