Multi-label stream data classification method based on Hough-ding tree

Abstract

The invention discloses a multi-label stream data classification method based on a Hough-ding tree, and belongs to the technical field of pattern recognition and data mining. A training stage step: enabling new examples to flow into the model one by one; constructing a base classifier according to the number of instance features, respectively distributing a plurality of features, and carrying out incremental learning; the structures are used as basic units for hierarchical cascading, and each layer is provided with a plurality of different basic units; performing three-fold cross validation during training of each layer to obtain the weight of each unit, and verifying the classification performance of the current layer; a prediction stage step: enabling the new examples to flow into the trained model one by one, and obtaining the feature vector of the current example; feature division; the weight determined when training is applied to the predicted value of the current instance by each basic unit is spliced with the original feature to serve as the input feature of the next layer; and performing statistics on the prediction precision of each layer for the current test instance, if the prediction precision does not obviously increase, stopping model increase, and taking the prediction value of the last layer as a final prediction result.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and data mining, and in particular relates to a method for classifying multi-label flow data based on a Hofferding tree. Background technique [0002] With the popularization and vigorous development of computer technology, data information is becoming more and more complex and changeable. How to filter useful information from massive data information is a key problem that we need to solve urgently in the era of big data. Feature selection, as a feature dimensionality reduction technique that obtains an "optimal feature subset" by eliminating irrelevant and redundant features, has received extensive attention from researchers in the context of big data. [0003] Classification and regression problems are hot topics in machine learning. Existing multi-label classification methods usually deal with existing complete static data and put them into the model in batches in the form of featur...

AI Technical Summary

Problems solved by technology

[0005] Due to the limitations and one-sidedness of the technology for classification and label prediction of multi-label and streaming data under the existing big data background, as well as the problems of poor prediction performance and high time complexity

In order to solve this problem, the present invention provides a multi-label flow data classification method based on Houghting tree, in order to solve the label prediction problem of multi-label and flow data at the same time, and further improve the algorithm by using the cascade structure and hierarchical model performance, which can effectively improve the real-time performance of dynamic data flow label prediction

Images