Data feature selection method based on artificial bee colony algorithm

Abstract

The invention relates to a data feature selection method based on an artificial bee colony algorithm. The method comprises the steps that the control parameter of the artificial bee colony algorithm is determined, and appointment normalization processing is carried out on an acquired data set; a set of bee position is initialized; a fitness function is selected according to selected features to calculate the fitness value of each bee, and the corresponding mining number of times is set to zero; based on the update method of the artificial bee colony algorithm, the position of the bees is updated, and the fitness value of a new individual is calculated and the number of times of mining is updated; a probability model function is calculated and selected; a bee is selected as an observation bee; the position of the observation bee is updated; the fitness value of a new individual is calculated and the number of times of mining is updated; the number of times of mining is observed, and a bee position update mechanism is carried out; the optimal solution position so far is kept, and represents the optimal feature subset; and if the maximum number of times of iteration is up, the optimal feature subset is output, otherwise the steps are repeated. According to the invention, the complexity of a feature selection method can be reduced.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a data feature selection method based on an artificial bee colony algorithm. Background technique [0002] In recent years, data mining technology has been widely used in the fields of business intelligence, biomedicine and genetic testing, and how to reduce the dimensionality from large-scale data to obtain effective simplified data is becoming more and more important. In many practical applications, data sets stored in databases often have thousands or even tens of thousands of features, but not all features are helpful for discovering important information hidden behind the data. Since only a small part of the features represent the distribution characteristics of the entire high-dimensional feature space, these features that need to be deleted not only increase the interference in the process of knowledge discovery by the learning algorithm, but also increase the comp...

AI Technical Summary

Problems solved by technology

However, this is impossible for large datasets with many features, so this method is rarely used in practical applications

In addition, when using the encapsulation formula to select the optimal feature subset, the minimization of the optimal feature number is not considered

Images