基于机器学习的酒精偏好者特征分析及酗酒倾向预测

Analysis of Alcohol PreferenceCharacteristics and Prediction of Alcohol Abuse Tendency Based on Machine Learning

摘要: 饮酒超出适量饮酒或一般社交性饮酒的标准为重度饮酒，重度饮酒无论对个人发展还是对家庭、社会都会产生严重的负面影响，因此，通过特征进行酗酒原因分析和行为预测具有重要意义。特征包括基本信息(教育水平、年龄、性别、居住国、种族)以及五因素人格测量(神经质、外向性、开放性、友善性、严谨性)，还有巴瑞特冲动性和冲动感觉寻求，基于机器学习的决策树、朴素贝叶斯、K近邻、支持向量机、逻辑回归多种分类方法进行预测，通过对一个人的基本信息和性格特征的分析来预测是否具有酗酒倾向。对酗酒者的特征进行了分析，外向对饮酒行为有较大影响，开放性高的人更不倾向于饮酒，在其他性格特征中，神经质，友善性，严谨性，巴瑞特冲动性和感觉寻求越高，饮酒可能性越大。

Abstract: Heavy drinking refers to drinking exceeds the standard of moderate drinking or general social drinking. Heavy drinking has serious negative effects on personal development and family, so the analysis of the causes of alcohol abuse and the behavior prediction of heavy drinking are important. Features include basic information (educational level, age, gender, country of residence, ethnicity) and five-factor personality measures (Neuroticism, Extraversion, Openness to experience, Agreeableness, Conscientiousness) as well as Barrett impulsiveness and impulsive sensations seeking. We used machine learning-based decision trees, Naive Bayes, K-nearest neighbors, support vector machines, logistic regression and other classification methods to predict and analyze. Then we can predict whether there is a tendency to alcohol abuse according to a person’s basic information and personality characteristics. And the characteristics of alcoholics were analyzed: extroversion had a great influence on drinking behavior; people with high openness were less inclined to drinking alcohol, as for other personality traits, the higher Neuroticism, Agreeableness, Conscientiousness, Barrett impulsiveness and feeling seeking, the greater the possibility of drinking.