USING FUZZY TIME-SERIES PREDICTION FOR FORECASTING RAINFALL IN TAIWAN

Forecasting is an important academic research topic and method. This method is widely applied in various research fields, such as research fields related to industry, commerce, agriculture, economics, medicine, the environment, social science, and engineering. Scholars have proposed different analytical methods for conducting forecasting research. The proposed methods include quantitative methods, such as questionnaire or field surveys, which provide statistical data, and qualitative methods, such as interviews or observations, which provide analytical data. However, considerable labor, material resources, and time are required to acquire the necessary quantitative statistical data or qualitative analytical data for conducting forecasting research. In the era of globalization and rapidly changing virtual networks, forecast research is crucial. Therefore, rapidly obtaining research data and applying soft computing and effective prediction tools are essential. In this study, the average annual rainfall in Taiwan for 15 years (2005-2019), which is recorded in the statistical database of the Taiwanese government's official website, was used as an example for forecasting. Soft computing involving fuzzy time-series prediction was used to examine empirically the applicability of a fuzzy time-series prediction model for the prediction of rainfall data. Only 15 historical time-series data points were used to perform the prediction. The use of 10-14 fuzzy membership subsets yielded a prediction error of less than 8%, which indicated that a small amount of data can be used to obtain favorable forecasting results with the adopted method.