Random forest training method and device, storage medium and electronic equipment

Abstract

The invention relates to a random forest training method and device, a storage medium and electronic equipment. The method comprises: determining n groups of training data sets in first training data;Judging the trained n trees through the description data of the first training data, and obtaining n prediction results; Deleting the n trees according to the correct rates of the n prediction results and a preset threshold value to obtain m trees; Voting the m trees according to the weight corresponding to each tree in the m trees to obtain a target tree; Synthesizing the prediction result corresponding to the target tree and the description data into second training data; And taking the second training data as the first training data, and circularly executing the above steps until the correct rate of the n prediction results is greater than or equal to the preset threshold, and obtaining the random forest. According to the method, the overall training data can be continuously optimizedin the multiple training processes of the random forest, so that the accuracy of classification prediction is improved while increasing of trees with single characteristics in the training process isavoided.

Description

technical field [0001] The present disclosure relates to the field of machine learning, in particular, to a random forest training method, device, storage medium and electronic equipment. Background technique [0002] A random forest is a classifier that contains multiple decision trees, and its output prediction is determined by the mode of the prediction output by each tree. The decision tree is a tree-structured model for supervised learning. In supervised learning, a set of samples can be given first, each sample contains a set of attributes (descriptive data) and a category (prediction result), these categories are determined in advance, by learning this set of samples can get a The decision tree of the classification function, which can give the correct classification (output prediction result) to the new object. In related technologies, when training a random forest, each decision tree in the random forest is usually trained once with a part of the full training dat...

AI Technical Summary

Problems solved by technology

However, the prediction accuracy rate of the decision tree that only undergoes a single training is not high, and it cannot cope with the unbalanced data characteristics in the training data during the training process (there is a lot of data in a certain category), which in turn reduces the accuracy of the entire classification prediction. question

Images