Fuzzy-Autoregressive Integrated Moving Average (F-ARIMA) Model to Improve Temperature Forecast

Abstract

A temperature forecast is a form of weather forecast that predicts temperature conditions using science and technology. Temperature forecasting is critical for making decisions in a variety of activities. To attain high predicted accuracy, predictive models must be built using accurate historical data. Data obtained through multiple methods, on the other hand, is inherently unreliable, resulting in less reliable predictive models. Hence, the data must be carefully managed, particularly to remove data uncertainty. While traditional data processing systems are simple to employ, they lack standard approaches for dealing with data uncertainty. As a consequence, this research presents a method for predicting temperature using ARIMA, as well as fuzzy data preparation strategies for dealing with fuzzy data during the data pre-processing phase. Standard deviation approaches were used to build fuzzy triangles for managing fuzzy data. The proposed method for creating fuzzy numbers using standard deviations yields fewer prediction errors and increases model performance, according to the experimental results. This is because data errors have been rectified, and model development errors have been decreased.