Multilayer Perceptron Neural Network Based Daily Temperature Forecasting for Kashan Using Meteorological Data from ۲۰۰۶ to ۲۰۲۵

Accurate short term temperature forecasting is important for agriculture, energy management, water resource planning, and public health. This need is even greater in arid and semi arid regions such as central Iran. This paper presents a data driven approach for forecasting daily air temperature in the city of Kashan. The approach uses an Artificial Neural Network (ANN) based on the Multilayer Perceptron (MLP) architecture. A meteorological dataset that covers the period from ۲۰۰۶ to ۲۰۲۵ was collected, preprocessed, and normalized. This dataset was then used to train and validate the proposed model. The network was trained with the Levenberg Marquardt back propagation algorithm. Several input configurations and hidden layer sizes were examined in order to obtain the optimal structure. The performance of the model was evaluated using three standard statistical metrics. These metrics are the Root Mean Square Error (RMSE), the Mean Absolute Error (MAE), and the coefficient of determination (R²). The experimental results show that the proposed MLP network can capture the nonlinear behavior of the regional climate. The model also provides accurate temperature predictions. These findings confirm that neural network models are suitable for meteorological forecasting in arid regions.

کلیدواژه ها:

Temperature forecasting ، Artificial Neural Network ، Multilayer Perceptron ، Levenberg Marquardt ، Kashan ، Time series prediction