基于Xgboost和Keras框架的多疾病风险预测

Multi-disease risk prediction based on Xgboost and Keras frameworks

慢性病已成为人类健康的主要威胁,找出慢性病发生直接的或间接的因素,做好疾病风险预测有着重要意义。本文采用3种集成学习算法RF、GBDT和Xgboost对3种慢性病进行分类,采用分类效果最好的Xgboost进行特征选择,使用Keras深度学习框架构建神经网络进行多疾病风险预测,采用问题转化中BR和LP二种方法将多疾病风险预测转化为多标签分类问题。

Chronic diseases have become a major threat to human health.It is important to find out the direct or indirect factors of chronic diseases to predict the risk of multiple diseases.In this paper,RF,GBDT and Xgboost are used to classify the three chronic diseases,and the Xgboost with the best classification effect is used for feature selection.The Keras deep learning framework is used to construct a neural network to predict the risk of multiple diseases.The multiple disease risk prediction is transformed into multi label classification problem by using Binary Relevance(BR)and Label Powerset(LP)in problem transformation.