A Systematic Sample Size Analysis of ML-Based Path Loss Prediction for LPWAN

2026-08-11

AI Summary

This study addresses the limited accuracy of path loss prediction for low-power wide-area networks such as LoRa in urban environments by systematically quantifying the impact of training sample size on machine learning model performance. Leveraging real-world LoRa deployment data augmented with LiDAR-derived terrain features and geographic coordinates, the authors employ random forest and k-nearest neighbors algorithms to model path loss, evaluating both interpolation and cross-gateway extrapolation scenarios via leave-one-gateway-out cross-validation. Results demonstrate that, with the largest training set, the models achieve a root mean square error (RMSE) as low as 6.5 dB—significantly outperforming the best baseline (9.7 dB)—though performance degrades when extrapolating to unseen gateways. This work is the first to reveal the critical relationship between training data scale and generalization capability in practical LoRa deployments.

Abstract

Low Power Wide Area Networks like LoRa are increasingly deployed for smart city applications, requiring accurate path loss prediction for effective network planning. Traditional (empirical) propagation models often exhibit limited accuracy in these scenarios. We investigate machine learning models for LoRa path loss prediction, systematically analyzing how prediction accuracy scales with training set size using real-world measurements from an urban deployment. Our approach employs a Random Forest with LiDAR-derived terrain features and k-Nearest Neighbors with coordinate data, comparing their performance against established empirical models and specialized LPWAN models. Under random pooled splits, both ML models consistently outperform the considered baseline models across the evaluated training-set sizes. At maximum training size, they achieve RMSE values below 6.5 dB compared to 9.7 dB for the best baseline, indicating accurate within-deployment interpolation. A leave-one-gateway-out check qualifies this result: RF shows placement-dependent transfer to held-out gateways, with moderate degradation for several gateways but larger errors for others, whereas coordinate-only k-NN degrades substantially when the gateway location is unseen

Problem

• path loss prediction
• LPWAN
• LoRa
• network planning
• propagation models

Innovation

• machine learning
• path loss prediction
• LPWAN
• sample size analysis
• LiDAR features