Monitoring the mental workload of construction workers is effective in detecting risky subjects because cognitive overload may threaten their safety. This study aimed to measure workers’ mental workload caused by heat exposure using heart rate variability (HRV) and eye movement features. Inexperienced pipe workers (n = 30) were invited to perform an installation task in a normothermic (26 °C, 50% RH) and a hyperthermic (33 °C, 50% RH) condition. Their HRV and eye movement features were recorded as the inputs of training models classifying mental workload between the two thermal conditions, using supervised machine learning algorithms, including Support Vector Machines (SVM), KNearest Neighbor (KNN), Linear Discriminant Analysis (LDA), and Random Forest (RF). The results show that applying eight HRV features through the KNN algorithm could obtain the highest classification accuracy of 90.00% (Recall = 0.933, Precision = 0.875, F1 = 0.903, AUC = 0.887). This study could provide a new perspective for monitoring the mental workload of construction workers, and it could also provide a feasible approach for the construction industry to monitor workers’ mental workload in hot conditions.