<![CDATA[The low accuracy of conventional weather prediction models is often hampered by the complexity of weather data that is dynamic, non-linear, and full of uncertainty. Therefore, it is necessary to evaluate hybrid models such as KNN-Naïve Bayes with cross-validation techniques to test their reliability and effectiveness in producing more accurate and consistent weather predictions. This study evaluates the effectiveness of the hybrid KNN-Naïve Bayes model for weather prediction using cross-validation techniques. Through an experimental quantitative approach, weather data from Kaggle was analyzed to compare the performance of the KNN, Naïve Bayes, and hybrid KNN-Naïve Bayes models. Data preprocessing methods include normalization and dataset partitioning with five variations of data sharing ratios. Model evaluation using 3-fold cross-validation shows that the hybrid KNN-Naïve Bayes model achieves the highest accuracy of 97.20% at ratios of 80:20 and 70:30, outperforming KNN with an accuracy of 96.27% and Naïve Bayes with 96.56%. The implications of this research indicate that the hybrid model can overcome the limitations of each algorithm, particularly in handling class imbalance and the assumption of feature independence. The results of the KNN-Naïve Bayes hybrid model research proved to be a superior alternative in weather prediction that can contribute to the development of a more reliable early warning system in the face of increasingly dynamic climate change. In conclusion, the application of the KNN-Naïve Bayes hybrid model with cross-validation techniques can improve the accuracy of weather prediction compared to the use of either method separately.]]>
Copyrights © 2025
