Solar radiation forecasting is important for energy management, energy stability and installation of solar energy systems. In this study, hybrid LSTM-SVM was developed to estimate the daily solar radiation of Nusaybin district of Mardin province. To compare the hybrid LSTM-SVM, support vector machines (SVM), long short-term memory (LSTM), decision tree and K-nearest neighbor were used. The performance of the models was evaluated with the help of coefficient of determination ( \({R}^{2}\) ), mean square error (MSE), root-mean-square error (RMSE), normalized root-mean-square error (NRMSE), mean absolute percentage error (MAPE) and Taylor diagram. The regression curve of the models was drawn and their solar radiation prediction performances were compared. The correlation matrix determined the correlation between data. \({R}^{2}\) , MSE, RMSE, NRMSE and MAPE of the hybrid LSTM-SVM were 0.962, 2.0191e-04, 0.0142, 0.0384 and 42.7962, respectively. The proposed hybrid LSTM-SVM had higher performance than other compared models. Taylor diagram proved that the prediction accuracy of the hybrid LSTM-SVM model is good. A correct correlation of 0.8315 between solar radiation and sunshine duration was determined by the correlation matrix. The meteorological data that most affected solar radiation was sunshine duration. As a result, the hybrid LSTM-SVM can be used as an alternative method for solar radiation estimation.
