Forecasting short-term indoor radon: a machine learning approach using LSTM networks

Abstract

Indoor radon is a radioactive gas that can accumulate in homes and pose a health risk for humans. Forecasting indoor radon levels may be used as a tool for mitigating human exposure risk, and thus help to effectively manage indoor radon risk. Forecasting based on Machine Learning (ML) techniques involves predicting future levels of indoor radon gas based on past and current data, and thus help identify trends and patterns in the data over time. This work presents preliminary results regarding the implementation and evaluation of two LSTMbased approaches, for indoor radon forecasting, which can then be used as a tool to trigger preventive management procedures for Indoor Air Quality management. Preliminary results have shown that the normalized data using the Long Short-Term Memory (LSTM) algorithm proved to be the optimal approach for this application case, demonstrating superior accuracy across various forecasting time windows when compared to other approaches evaluated in this work.