Akademik Veri Yönetim Sistemi
6. INTERNATIONAL MEDITERRANEAN CONGRESS, Rome, İtalya, 13 - 15 Ağustos 2024, cilt.1, sa.1, ss.654-659
Air temperature is expressed as the measurement of the temperature or coldness of the surrounding air, which varies according to the location, time of day, and weather conditions. In recent years, air temperature forecasting has become increasingly important for countries engaged in agriculture. Constantly changing weather conditions make air temperature forecasting more complex. Today, various statistical models have begun to be used for air temperature forecasting. Observation values obtained within a specific time interval are referred to as time series. Time series that include discrete, linear, and stochastic processes are called ARIMA models. In an ARIMA model, the dependent variable is explained by its lagged values and random error terms. The goal of ARIMA models is to identify the linear model that fits the time series best and contains the fewest parameters. In time series data, there may be values that are significantly larger or smaller than other data points, indicating different behavior. These values are called outliers. When outliers are present in ARIMA models, the obtained estimates and forecast values may be unreliable; instead, alternative Robust ARIMA models are used. In the literature, these are known as REGARIMA models.
In this study, it is aimed to apply various time series and alternative time series analysis methods for air temperature forecasting and to compare their performances. For this purpose, using air temperature data belonging to Istanbul Province, future forecast values were obtained by using appropriate ARIMA and Robust ARIMA models for the relevant time series. In the next stage, which technique gives better results is discussed using different comparison criteria