Increasing Prediction Accuracy using Pca-Regression Technique in Ehwsn

Energy Harvested Wireless Sensor Networks (EHWSN) play a vital role in mission-critical applications such as forest fire detection and industrial control and safety applications. Accurate solar forecasting is essential for the proper design of the Medium Access Control protocol (MAC). Recent work has focused on considering only a single variable, i.e., previous solar data, for solar forecasting, resulting in low accuracy. In the proposed work, covariance among the features was determined using Principal Component Analysis (PCA). Only those features showing strong correlations were retained. Subsequently, the KNN-Regression technique, applied with the reduced feature set, demonstrated a significant improvement over the previous method. The proposed approach achieved an accuracy of 92.47% with a Root Mean Square Error (RMSE) of 11.52, compared to the previous figures of 85.14 and 17.11, respectively.