Purpose/Significance The internet consultation user portrait is constructed to explore the consultation topic, improve the consultation service quality, reduce the communication barriers between doctors and patients, and provide targeted treatment in an online and offline manner. Method/Process Python crawler is used to obtain the autism diagnosis data of a medical platform, and the combined model of LDA and TF-TFIDF is used to divide the data, and the user group classification is realized after dimensionality reduction clustering. Finally, the characteristic sets of different user groups are calculated and output by logistic regression model to construct the portrait. Result/Conclusion The consultation content of users mainly focuses on 11 topics. The platform can optimize the consultation filling template based on the typical characteristics of the subject content to improve the accuracy of the disease description, consultation efficiency and satisfaction of patients.