Predicting risky sexual behavior among college students through machine learning approaches: Analysis of individual data from 1264 universities in 31 provinces in China (Preprint)

• DOI: 10.2196/preprints.41162
• Publication Date: 2022-07-20T19:38:10Z
• Authors: Xuan Li; Hanxiyue Zhang; Shuangyu Zhao; Kun Tang
• Keywords: Casual; Sexual intercourse

<sec> <title>BACKGROUND</title> Risky sexual behavior (RSB), as the most direct risk factor for sexually transmitted infections (STIs), is common among college students. Thus, it important to intervene and prevent students by identifying relevant factors making predictions. </sec> <title>OBJECTIVE</title> We aimed establish a predictive model RSB facilitate timely prevention intervention before contraction of STIs. <title>METHODS</title> included total 8,290 self-reported heterosexual Chinese with intercourse experience from November 2019 February 2020. identified those attributed four dimensions: whether contraception was used; contraceptive method safe; engaged in casual sex or multiple partners; integrated RSB, which combined first three dimensions. For each type, we compared various machine learning (ML) models according validation indicators chose optimal both prediction identification. <title>RESULTS</title> In total, 4993 (60·2%) had ever RSB. Among them, 3422 (41·3%) did not use every time they intercourse, 3393 (40·93%) used an unsafe method, 1069 (12·9%) partners. Through comparison, XGBoost (XGB) gradient boosting (GBM) achieved performance on area under receiver operator characteristic curve (AUC) reaching 0·80. Under condition ensuring stability indicators, 12 variables were finally selected XGB, including participants’ relationship status, knowledge, attitude, previous experience. <title>CONCLUSIONS</title> prevalent students, ML effective approach predict identify corresponding factors.