Building-related symptoms (BRS) such as headache, tiredness, dry skin, cough are a common issue in schools, with significant implications for students’ health and academic performance. This study aims to predict the prevalence of self-reported symptoms (SRS) using machine learning (ML) models based on environmental parameters and prevalence of indoor environmental complaints (IEC). Environmental parameters were collected through indoor sensors, while outdoor data represented a typical cold climate region. Additionally, student surveys gathered data on IEC and perceived SRS. The prevalence of indoor environmental complaints was calculated together with environmental parameters are used as in-puts, where prevalence of SRS as targets for ML modelling with three algorithms—Multi-linear Regression (MLR), Decision Tree (DT), and Random Forest (RF). Models incorporating prevalence of IEC as part of ML input significantly improved prediction accuracy (R2 testing results, up to 0.9) compared to models without prevalence of IEC (R2 < 0.6). RF provided the most accurate predictions with reduced overfitting compared to DT and higher accuracy than MLR. ML feature importance analysis revealed that the prevalence of IEC was the most influential predictor, with addition of some environmental parameters playing a complementary role. The findings emphasize the importance of IEC in under-standing the prevalence of SRS. Future research should investigate additional predictors to enhance the accuracy of ML models for prediction of prevalence of SRS.