Accuracy increase in forecasting process of meteorological phenomena, including dew point temperature due to freshwater shortage and the consequences of climate change, are significant topics. In this regard, the improved clustering algorithm was presented by implementing it in five meteorological stations of West Azerbaijan Province. At the algorithm implementation beginning, dew point temperature was determined by sixteen experimental models with different mathematical structures, and then the dew point temperature data were entered into the clustering process. Then, the new data obtained by averaging the data of each cluster were entered into the artificial intelligence model. Based on the evaluation statistics, two models M1 and M8 with temperature and relative humidity dependence and different mathematical structure had high accuracy. Some models, including the models of group seven with different mathematical structures, were highly sensitive to changes in coefficients. The performance of the extended clustering method increased significantly compared to the two superior models (the root mean square error reduction at all stations from M1 and M8 to the latter method was 59.71% and 66.31%, relative root mean square error reduction was 59.7% and 66.36%, mean absolute percentage square error reduction was 62.74% and 70.8%). The minimum error rate from least to greatest, includes Maku, Urmia, Khoy, Mahabad, and Salmas. The improved clustering method had an overestimation except Salmas station. The research algorithm was able to greatly increase the dew point temperature estimation accuracy by operating different models with different efficiencies and classifying their data and extracting information correctly.