Uncertain data classification method based on extreme learning machine

Abstract

The invention discloses an uncertain data classification method based on an extreme learning machine. In an uncertain extreme learning machine (UELM) method, modeling is performed on uncertain data by use of a probability density function, so, a problem of a probability distribution information loss caused by an expectation value and sampling point-based method is effectively avoided. In addition, in the UELM method, a traditional extreme learning machine method framework is re-designed and reception data of an input layer and activation functions of a hidden layer are modified, so through this improvement, the extreme learning machine is allowed to be quite suitable for uncertain data. The whole UELM method is divided into an initialization stage, a training stage and a prediction stage. In the initialization stage, experiment parameters are generated. In the training stage, learning results are obtained through a learning process of experiment data. In the prediction stage, new data is classified by use of results which is obtained through learning in the training stage. A few of experiment results show that the UELM method performs brilliantly in accuracy and time efficiency compared with other uncertain data classification methods.

Description

technical field [0001] The invention belongs to the technical field of uncertain data classification in the academic field, and in particular relates to an uncertain data classification method based on an extreme learning machine. Background technique [0002] Existing classification methods are widely used in numerically unique data classification problems. However, in many real applications in the fields of military industry, location-based services, finance, etc., uncertain data whose values ​​follow a certain distribution are ubiquitous. If traditional classification methods are used to classify uncertain data, it is very likely to give Wrong results, wrong information to decision makers, resulting in losses. Therefore, how to correctly classify uncertain data has extremely important research value for practical applications. The mainstream classification methods such as decision tree, SVM, and nearest neighbor are too sensitive to uncertain data and cannot meet the re...

AI Technical Summary

Problems solved by technology

The mainstream classification methods such as decision tree, SVM, and nearest neighbor are too sensitive to uncertain data, and cannot meet the requirements for classification accuracy and time efficiency of uncertain data. Further exploration of more accurate and efficient uncertain data classification methods is needed.

Images