Weather forecasting is important for the survival of the wider community. Therefore, the accuracy of the weather forecast must be high. Based on this, a study was conducted to improve the accuracy of weather forecasting with the naïve Bayes and C4.5 models and then performed an optimization using the ensemble method. The dataset used is weather data observed from BMKG Bandung for 10 years. Accuracy in the pretest process shows that the naïve Bayes algorithm has an accuracy of 49.45% and the C4.5 algorithm produces 41.24% accuracy, while in the posttest process the accuracy obtained is 49.76% for bagging naïve Bayes, 46.47% for boosting naïve Bayes, 45.76 for bagging C4.5 and 38.82% for C4.5.