Feature-weight-based LARS diabetes prediction method

Abstract

The invention relates to the technical field of medical informatization, and discloses a feature-weight-based LARS diabetes prediction method. Specifically the feature-weight-based LARS diabetes prediction method comprises the following steps: 1) normalizing a diabetes data set matrix, and initializing a current fitting value vector and a residual vector; 2) calculating an independent variable feature weight vector and an original relevancy vector; 3) calculating a unit vector, a regression coefficient vector, a new relevancy vector and the maximum relevancy; 4) updating a regression coefficient vector, a fitting value vector, a residual vector and an index set; and 5) judging whether the L2 norm of the residual vector is smaller than tolerance or not, and if so, ending, and if not, repeating the step 3 to the step 5. According to the feature-weight-based LARS diabetes prediction method, starting from the features of the diabetes data set, key feature variables of diabetes are screenedout, and a diabetes prediction model is simplified; and the accuracy of the diabetes prediction model is improved, so that accurate diabetes prevention and treatment measures can be provided.

Description

technical field [0001] The invention relates to the technical field of medical information technology, in particular to a LARS diabetes prediction method based on feature weights. Background technique [0002] As the data features of the diabetes prediction model become more and more, the data dimension becomes larger and larger, and the prediction model becomes more and more complex. It is difficult for traditional prediction methods to be directly applied to the prediction of diabetes. [0003] The increase in data features and data dimensions increases the training time complexity of the neural network prediction model, reduces the prediction accuracy and generalization ability of the decision tree and logistic regression prediction models, and makes it difficult for the support vector machine prediction model to directly find key features. Diabetes prediction models present new challenges. [0004] The lasso model has the advantages of high regression classification acc...

AI Technical Summary

Problems solved by technology

[0002] As the data features of the diabetes prediction model become more and more, the data dimension becomes larger and larger, and the prediction model becomes more and more complex. It is difficult for traditional prediction methods to be directly applied to the prediction of diabetes.

[0003] The increase in data features and data dimensions increases the training time complexity of the neural network prediction model, reduces the prediction accuracy and generalization ability of the decision tree and logistic regression prediction models, and makes it difficult for the support vector machine prediction model to directly find key features. Diabetes prediction models present new challenges

[0004] The lasso model has the advantages of high regression classification accuracy and strong generalization ability, but because the traditional minimum angle regression LARS algorithm has the problem of slow approximation speed and low accuracy when solving Lasso regression coefficients, it is difficult to use LARS algorithm in diabetes prediction

Images