Energy efficiency has become a hot topic in the last years since its relationship with energy consumption, climate change and the actual limitation of natural resources. Buildings account a significant proportion of said consumption and, in particular, their conditioning systems, due to its connection with indoor thermal comfort. Although in building heating, ventilation and air-conditioning field, efforts have been made on predicting building occupants’ thermal comfort, fewer studies face the analysis of the adaptive behaviour of the occupants, which plays a fundamental role in their thermal comfort. Based on that, in the present study, a machine learning approach is proposed for predicting the indoor clothing insulation level. Four algorithms were selected and compared for a field study database. The results show that multilayer perceptron neural network achieved better clothing insulation level prediction accuracy than others and that the most relevant input variables for such predictions were indoor environmental variables.