Research And Application Of Medical Data Mining Based On Distance Metric Learning

With the emergence and popularity of intelligent technology equipment like medical wearable devices,medical health data grow rapidly and people care more about their own physical health status.However,it is unavoidable for us to be plagued by serious diseases unknowingly.More and more research on artificial intelligence and big data technology in the field of medical health offer the chance to explore the details hidden behind big data and make possible that people will get better health care service and stay away from diseases.Cancer and cardiovascular disease not only bring damage to human body,but also put people under huge pressure,so it is necessary to use data mining technology for early diagnosis and prevention of diseases.Therefore,based on medical health examination data from physical examination institution in real life,a new distance metric algorithm adopting cosine metric and LMNN named COS-SUBLMNN is proposed to improve the classification accuracy and then this paper constructs the model of risk screening of major diseases and identification of high risk population combining with the feedback of patient's diagnosis.The main contents of this paper are as follows:studying the distance metric learning algorithm mainly,constructing the model of risk screening of cancer,the evaluation and optimization of algorithm and the construction of system of early diagnosis of major diseases.Firstly,COS-SUBLMNN algorithm is proposed adopting cosine metric and LMNN to improve the classification accuracy;and then this paper designs the solution of the preprocess of medical health examination data and feature selection,constructs the model based on LMNN in the application of early screening for cancer compared with some machine learning algorithm and choose criteria to evaluate model and make some optimization.Finally,this paper designs the system of early diagnosis of major diseases using COS-SUBLMNN with the feedback of patient's diagnosis verified by cancer and cardiovascular disease.The contribution of this paper is to propose COS-SUBLMNN algorithm and design the system of early diagnosis of major diseases based on distance metric learning.We verifies that LMNN and SVM,an algorithm with the idea of distance metric performs better than RF and ANN on the same sample set in risk screening of cancer and COS-LMNN provides an effective way of early diagnosis of cancer and cardiovascular disease and distance metric learning algorithm achieve the goal of disease diagnosis and identification of high risk population in the medical health examination data.The results of this study have practical significance in the field of medical health.