Delegated Regressor, A Robust Approach for Automated Anomaly Detection in the Soil Radon Time Series Data.

Area of Science:

• Environmental Science
• Geophysics
• Data Science

Background:

• Soil radon gas concentration (SRGC) is a crucial environmental indicator.
• Predicting SRGC and its anomalies is vital for environmental monitoring and hazard assessment.
• Existing methods for time series prediction have limitations in accuracy.

Purpose of the Study:

• To introduce a new delegating regressors method for predicting SRGC.
• To compare the proposed method's accuracy against traditional boosting and regression techniques.
• To identify optimal parameters for accurate SRGC prediction.

Main Methods:

• Development of a novel delegating regressors methodology.
• Comparative analysis using Extreme Gradient Boosting (EGB) and Support Vector Regressors (SVR).
• Statistical analysis of radon time series (RTS) data using R language.

Main Results:

• The proposed delegating regressors method demonstrated superior accuracy in predicting SRGC.
• The highest correlation between actual and predicted SRGC was observed with a window size of 2 days.
• The methodology showed precise predictions across various window sizes by overlapping predicted and actual radon time series.

Conclusions:

• The delegating regressors method offers a more accurate approach to SRGC prediction.
• The findings highlight the method's effectiveness in environmental time series analysis.
• The study provides a valuable tool for monitoring and understanding radon gas dynamics.