In order to further improve the intelligence and comfort level of the thermal system of automobile cabins, a user portrait-based cabin thermal comfort temperature decision algorithm is proposed. The age, gender, body mass index (BMI), cloth class and height features of users are obtained by designing three neural networks. Subsequently, the predicted percentage of dissatisfaction (PPD) index is calculated according to the obtained 5 features. With using the iteration optimization strategy, the cabin thermal comfort temperature is generated with satisfying the ISO7730 standard. Experimental test results show that the proposed method can efficiently reduce the PDD index so as to satisfy the thermal comfort, revealing the application value of the user portrait-based decision strategy.