基于SMOTEENN算法的音乐APP听歌习惯与大学生抑郁倾向的相关性研究  

Correlation between music APP listening habits and depression tendency in college students based on SMOTEENN algorithm

作　　者：黄馨巧 朱珲 屈壕 伍亚舟[1] 宋秋月[1] HUANG Xinqiao;ZHU Hui;QU Hao;WU Yazhou;SONG Qiuyue(Department of Health Statistics,Faculty of Military Preventive Medicine,Army Medical University(Third Military Medical University),Chongqing,400038,China)

机构地区：[1]陆军军医大学(第三军医大学)军事预防医学系军队卫生统计学教研室,重庆400038

出　　处：《陆军军医大学学报》2024年第23期2670-2680,共11页Journal of Army Medical University

基　　金：国家自然科学基金面上项目(82173621)。

摘　　要：目的 探究基于音乐APP听歌习惯的大学生抑郁倾向影响因素,构建预测模型并进一步优化。方法 2023年4~5月采用方便抽样对1 157份在校大学生进行问卷调查,利用单因素分析和Logistic回归分析筛选影响因素,在此基础上构建预测模型;采用SMOTEENN过抽样算法改进数据集,构建预测模型。结果 Logistic回归分析发现性别:女性(OR=1.730,95%CI:1.257~2.396),年级:大四和研究生(OR=2.649,95%CI:1.198~7.506)、(OR=2.041,95%CI:1.231~3.885),专业:理科(OR=1.573,95%CI:1.052~2.350),每天听歌时长:0.5~2 h(OR=1.661,95%CI:1.011~2.695),听歌曲风:“伤情”和“怀旧”(OR=2.668,95%CI:1.701~4.226)、(OR=1.751,95%CI:1.086~2.837),留言频率:“0~5%”的歌曲留言(OR=2.938,95%CI:1.018~8.417)是抑郁倾向的独立危险因素。开始听歌的年限:1~3年内(OR=0.547,95%CI:0.347~0.872),听歌时间段:14:00~18:00和18:00~21:00(OR=0.375,95%CI:0.167~0.845)、(OR=0.313,95%CI:0.148~0.671),“国风”类歌曲:喜欢(OR=0.711,95%CI:0.541~0.941)是抑郁倾向的独立保护因素。预测效果最优的模型为基于SMOTEENN算法的Logistic预警模型,AUC为0.923。结论 构建Logistic回归模型得到大学生抑郁倾向的9个独立影响因素。基于SMOTEENN算法所构建的预警模型能更准确地预测大学生的抑郁倾向。Objective To investigate the influencing factors for tendency towards depression in college students having music listening habits with music APP,and develop a prediction model and further optimize it.Methods A total of 1 157 college students were subjected with convenient sampling and surveyed with questionaires between April and May 2023.Univariate analysis and logistic regression analysis were employed to identify the influencing factors.Then a prediction model was constructed based on these factors.SMOTEENN over-sampling algorithm was utilized to enhance the dataset and construct the prediction model.Results Logistic regression analysis revealed that female(OR=1.730,95%CI:1.257~2.396),senior grade(OR=2.649,95%CI:1.198~7.506),postgraduate grade(OR=2.041,95%CI:1.231~3.885),major in Science(OR=1.573,95%CI:1.052~2.350),listening for a duration of 0.5~2 h(OR=1.661,95%CI:1.011~2.695),music style of melancholy(OR=2.668,95%CI:1.701~4.226) and of nostalgia(OR=1.751,95%CI:1.086~2.837),and frequency of comments on 0~5% of songs(OR=2.938,95%CI:1.018~8.417) were independent risk factors for depressive tendency.Time since listening to music for 1~3 years(OR=0.547,95%CI:0.347~0.872),listening to music from 14:00 to 18:00(OR=0.375,95%CI:0.167~0.845) and 18:00 to 21:00(OR=0.313,95%CI:0.148~0.671),and preference for Chinese style songs(OR=0.711,95%CI:0.541~0.941) were independent protective factors.The logistic early warning model based on SMOTEENN algorithm demonstrated optimal predictive performance with an AUC value of 0.923.Conclusion Our constructed logistic regression model has identified 9 independent influencing factors associated with depression tendency among college students.The early warning model based on SMOTEENN algorithm can predict the depression tendency more accurately for college students.

关 键 词：大学生 音乐APP 听歌习惯 抑郁倾向 预警模型 

分 类 号：G645.5[文化科学—高等教育学] R195.1[文化科学—教育学] R395.6[医药卫生—卫生统计学]