Regression tree prediction method, control device and computer readable storage medium

Abstract

The invention provides a regression tree prediction method, a control device and a computer readable storage medium. The method comprises the following steps: inputting a training data set; training aregression tree by using the training data set; whether the number of leaf nodes of the regression tree obtained through training is larger than 3 or not is judged, if yes, Jams-Stein estimation is adopted for estimating predicted values of all the leaf nodes at the same time; if not, independently estimating the predicted value of each leaf node by adopting maximum likelihood estimation; and outputting the regression tree and the predicted value. Obtaining better average prediction performance on all data sets than a regression tree based on maximum likelihood estimation, i.e., a smaller mean square error in average meaning; performance better than that of an existing optimal regression tree prediction method, namely kernel regression, is obtained on part of data sets; while the performance is improved, the high efficiency of an original regression tree based on maximum likelihood estimation prediction is basically maintained, and the influence of the number of data set samples on the test time is small.

Description

technical field [0001] The present invention relates to the technical field of machine learning, in particular to a regression tree prediction method, a control device and a computer-readable storage medium. Background technique [0002] In the field of machine learning, regression is a common machine learning task that can be used in tasks that require the prediction of continuous values, such as predicting house prices, hospital mortality, athlete scores, etc. Common regression methods include linear regression, logistic regression, nearest neighbor regression, support vector regression, and decision tree regression. [0003] Decision tree is a machine learning algorithm widely used in both academia and industry. If the final predicted value is discrete, it can be used for classification tasks. At this time, decision tree is simply called classification tree; otherwise, if given If the predicted value is continuous, it can be used for regression tasks, and the decision tr...

AI Technical Summary

Problems solved by technology

[0013] (1) In the regression tree prediction stage, each leaf node is estimated independently, and all leaf nodes are not considered together;

[0014] (2) There is no guarantee of consistent performance improvement on different datasets;

[0015] (3) Performance improvement and computing efficiency cannot be achieved at the same time

Images