Method for providing feature prediction capability based on naive Bayesian model

Abstract

The invention discloses a method for providing a feature prediction capability based on a naive Bayesian model. The method comprises the following steps: step 1, performing feature preprocessing according to an acquisition data set in the applied technical field to generate a feature data set; 2, extracting parameters in the model, and performing parameter adjustment according to the feature dataset through a ten-fold cross validation method to generate hyper-parameter values; 3, setting hyper-parameter values of a Gaussian naive Bayesian model, fitting the feature data set by using a fit function of the hyper-parameter values, and meanwhile, obtaining a feature value prediction probability model; and step 4, describing the probability distribution condition of the eigenvalue prediction probability model in the form of a histogram, improving the feature prediction capability of different fields of applied calculation, medicine, energy and voice by using the posteriori probability, andaccurately and effectively describing the feature prediction capability of the applied related technical fields.

Description

technical field [0001] The invention belongs to the field of feature prediction and analysis in machine learning, and in particular relates to a method for improving feature prediction ability based on a naive Bayesian model. Background technique [0002] The detection of feature predictive ability is a key issue in machine learning. There are many methods in machine learning to detect feature predictive ability. They can be divided into two categories. The first type is based on feature importance to detect feature predictive ability. , this feature prediction method is widely used in feature selection, for example: filtering methods use different evaluation criteria to evaluate the importance of features; The regression model is used to measure the importance of features; the method based on information theory uses different heuristic filtering criteria to measure the importance of features. In addition, the random forest method also often appears in the feature importanc...

AI Technical Summary

Problems solved by technology

[0003] The above methods always summarize the predictive ability from the perspective of features. In fact, the predictive ability of features is usually related to the change of feature values.

Images