Comparative analysis of a hybrid KNN–SVM model versus individual ML techniques for solar power forecasting in Bangladesh

This paper assesses a cascade hybrid KNN-SVM. hourly forecasting of solar power at Bangladesh and makes comparisons with standalone K-Nearest Neighbors (KNN), Support.Long Short Term Memory (LSTM), and Vector Machine (SVM).The hybrid strategy KNN determines local neighborhoods after which a local SVM with RBF kernel-achieves the lowest. Both HTGSR and HTPEG have RMSE of 0.0059, which is better. all individual models. While LSTM excels in long sequence tasks, the hybrid balances instancebased learning with margin optimization, ideal for high climatic variability. Results support scalable integration of solar energy into Bangladesh’s grid for sustainable planning.