Signal strength estimation in radio frequency systems under varying operational and environmental factors

Signal strength is critical to wireless communication systems, including cellular networks, Wi-Fi, satellites, and the Internet of Things (IoT). Strong signals support higher data rates and SNR while reducing latency. Conversely, weak signals cause dropped calls and poor connectivity. This study estimates RF signal strength using Machine Learning (ML) models. The dataset used contains 27288 samples with 15 inputs and one output (signal strength). To increase the efficiency of prediction, a study has combined three major models: Adaptive Neuro-Fuzzy Inference System (ANFIS), Stochastic Forest (SF), and CATBoost Gradient Boosting Regression (CATR). Partial Reinforcement Optimizer (PRO) and Starfish Optimization Algorithm (SOA) optimize model parameters, while Isolation Forest detects outliers. Five-fold cross-validation and R², RMSE, U95, SI, and WAPE evaluate performance. ANOVA assesses sensitivity. The best model, SFSO, achieves RMSE 2.43, R² 0.99, WAPE 0.04, and the lowest U95 of 6.75, demonstrating superior accuracy and reliability. .