Application of Support Vector Regression to the Prediction of the Long-Term Impacts of Climate Change on the Moisture Performance of Wood Frame and Massive Timber Walls

The objective of this study was to explore the potential a machine learning algorithm, Support Vector Machine Regression (SVR), forecast long-term hygrothermal responses and moisture performance light wood frame massive timber walls. Hygrothermal simulations were performed using 31-year long series climate data in three cities across Canada. Then, first 5 years used each case train model, which then (temperature relative humidity) indicator (mold growth index) for remaining series. location interest exterior layer OSB cross-laminated wall wall, respectively. A sliding window approach incorporate dependence response on past climatic conditions, allowed SVR capture time, implicitly. variable selection Least Absolute Shrinkage Selection Operator, revealed wind-driven rain, humidity, temperature, direct radiation as most contributing variables. results show that can be effectively cases studied. In some cases, discrepancies observed due lack capturing full range variability variables during years.