Research On The Application Of Combination Feature Selection Algorithm Based On CNN In Medical Data

With the explosion of the information age and the continuous optimization of data collection technologies,new data are being produced in an explosive manner in all industries,including the medical industry.Today,almost every hospital or medical unit has set up its own medical system,standardizing all procedures to effectively monitor the status of patients and record all data.The number of recorded medical data increases exponentially.Therefore,it is of great practical significance to use data analysis technology and machine learning algorithm to train and learn valuable unknown information by taking the case records of past patients as samples and applying them to disease prediction.In this paper,a combinational feature selection algorithm based on Convolutional Neural Network(CNN)is proposed for medical datasets,,which is based on the characteristic engineering and classification model of the data analysis task.The algorithm firstly transforms the original low-dimensional feature vectors into high-dimensional feature matrix form through feature combination,then uses the convolutional neural network for automatic feature importance analysis to achieve the effect of feature selection,and finally obtains the prediction results through the classification model of the convolutional neural network.The process of feature combination can enlarge the original feature space,enrich the information and improve the accuracy of model prediction.The powerful computing power and outstanding recognition ability of convolutional neural network can not only further improve the prediction accuracy of the model,but also eliminate the time consumption caused by the dimension increase caused by feature combination.In order to study the application effect of this algorithm in medical data,a real medical dataset and three open medical datasets were selected to carry out experiments respectively.In order to verify the excellent performance of this algorithm more clearly,this paper also uses the traditional classification algorithm for comparative experiments.These experimental results show that,on the one hand,the algorithm can generally achieve high prediction accuracy on multiple medical datasets.On the other hand,in terms of model prediction accuracy and prediction time,this algorithm can achieve better results.