A Probabilistic Approach to Predicting Energy Consumption in Wireless Sensor Networks using a Random Forest and SVM-based Model
B Neeraja, Sonali R. Nandanwar, Amit Joshi, Dr. M. Vijaya Bhaskar, Dharamvir Dharamvir, R. Jamuna · 2025
A common configuration for WSN is clusters of low-power sensor nodes placed in geographically separated areas. A CH is an individual assigned to each cluster whose job it is to collect data from all of the nodes in the cluster and send it to a BS in the center. Since these sensor nodes are usually powered by non-replaceable batteries, energy consumption poses a key problem to WSNs. Consistent data transmission and network lifetime depend on efficient energy utilisation. Improving data reliability and transmission efficiency, this study presents a new model for energy consumption in WSN, which solves the problem of excessive power use. By analysing and decreasing similarity in sensor data, the approach incorporates data pretreatment to minimised collecting errors and data redundancy. In order to choose features that are useful for classification, the chi-square approach is used. What makes this model unique is its hybrid RF-SVM architecture, which combines the best features of the popular RF and SVM classifiers. The experimental findings show that this method has a prediction accuracy of 99.51%, which is better than the current models. Results like this show that the RF-SVM approach could greatly benefit WSNs in terms of energy savings and fault detection.