Research On Medical Auxiliary Diagnosis Based On Feature Selection Of Clinical Data

With the rise of the modern digital wave,the generation of medical data has a factual support for the decision-making and operation of the medical system,and can better improve the operational efficiency of the medical system,such as neural network.However,how to extract effective information from medical data and assist in decisionmaking and diagnosis,and how to trace,track and predict a certain disease based on medical data is still an important research direction in the current medical data field.In the current concept of precision medicine,in assisting diagnosis with clinical data,it is meaningful how to make the results obtained through the algorithm model match the medical knowledge and make an appropriate diagnosis result.Existing data analysis algorithms can obtain better prediction accuracy for medical data tasks to a certain extent,but lack of consideration of the effectiveness of input features,which makes it difficult to explain the meaning of the selected features,and at the same time The accuracy of the analysis of small samples of medical data is difficult to guarantee and other problems.In response to the above-mentioned problems,this paper proposes corresponding methods for improvement from the perspective of feature selection.The main work of this paper is as follows:(1)Feature selection.Aiming at the issue of the validity of features,taking into account the issues of dimensionality reduction for high-latitude features,at the macro level,by defining the meaning of feature validity,using the characteristics and relationships between the features themselves to make selections.Based on this idea,this paper proposes An improved genetic algorithm based on maximum correlation and minimum redundancy is proposed to realize the search and selection of feature space.Experimental results show that the model can effectively reduce the feature dimension while retaining a good amount of original information.(2)Data prediction.In order to better obtain the effective features and improve the overall prediction accuracy of the algorithm,it is necessary to fit the feature space and model space.In this paper,an improved sequence forward selection algorithm based on random forest is proposed to achieve the bidirectional fitting of feature space and model space.Experimental results show that the model can further reduce the number of feature dimensions,and at the same time can improve the prediction accuracy of the overall algorithm.(3)Medical data clinical assistance system.Based on the feature selection model and data prediction model proposed in this paper,a clinical assistance system for medical data is designed and implemented,and the final prediction results can be obtained through functional modules such as data import and parameter setting.