身体质量指数（BMI）作为一种衡量个体身体质量的指标，可以依据BMI指数对个人身体质量水平进行分类，分类的特征变量涵盖了饮食、运动、生活方式等方面。本文应用了多种统计学习的方法，对身体质量指数建立多分类模型。所用到的分类方法包括单一分类算法和集成学习算法，比如朴素贝叶斯、K-最近邻、决策树、随机森林和XGBoost分类。在模型建立前，本文主要采用V折交叉验证，对需要预先给定参数的模型进行调参；使用训练数据建立模型后，本文在预测集上计算了模型分类的混淆矩阵，以及准确率、F1分数、Kappa系数等相关指标，对各个模型的分类效果进行了评价与对比，K-最近邻和XGBoost方法的分类准确率、精确率和召回率均超过80%。另外，本文在使用决策树、随机森林和极致梯度提升树建模分析时，计算了特征变量在分类模型中的相对重要性，找到了会对身体质量水平造成较大影响的变量。

Body mass index (BMI), as a measurement index of individual body quality, can be used for multi-classification. The characteristic variables of the classification include diet, exercise, lifestyle and other aspects. This paper uses a variety of statistical learning methods to establish a multi-classification model of body mass index. The classification methods include single classification algorithms and ensemble learning algorithms, such as naive bayes, K-nearest neighbor, decision trees, random forests, and XGBoost classification. Before establishing the model, this paper mainly uses the method of V-fold cross-validation to adjust parameters in advance. After establishing the model with training data, this paper uses the prediction set to calculate confusion matrix, accuracy, F1-score, and kappa coefficient. As a result, the accuracy, precision, and recall of  K-nearest neighbor and XGBoost methods are all over 80%. In addition, when using decision tree, random forest and extreme gradient boosting for modeling and analysis, this paper calculates the importance of variables and finds the important features.