Abstract: To address the issues of high cost and difficulty in large-scale deployment of traditional electric field strength measurements, this paper proposes a machine learning prediction method based on multi-parameter signal collaboration from smartphones. We synchronized a smartphone with a spectrum analyzer to acquire multi-dimensional signal parameters—including reference signal received power (RSRP), reference signal received quality (RSRQ), signal-to-interference-plus-noise ratio (SINR), and received signal strength indicator (RSSI)—while simultaneously capturing the ground-truth spatial field strength. A multi-parameter feature set was constructed after data processing and feature engineering. Comparative experiments based on various tree models and ensemble learning show that the Stacking multi-parameter model reduces the mean absolute error (MAE) of prediction to 0.71 dB, which is approximately 62.4% lower than that of the best single-parameter model, while the coefficient of determination (R²) increases from 0.76 to 0.96. Parameter-combination ablation experiments show that different signal parameters have complementary effects, and the complete four-parameter fusion model achieves the best prediction performance. Compared with existing chine learning models, the proposed method further reduces the prediction error by 26.8%, effectively demonstrating the effectiveness and advantages of integrating multi-dimensional signal parameters for field strength prediction. It provides a new solution for building a wide-coverage, “crowdsourced” electromagnetic environment monitoring network.