Method and device for extracting dimensionality reduction features

Abstract

The present application discloses a method and device for extracting dimensionality reduction features. The method includes: extracting a feature importance value in a preset scene from the training data, wherein the training data is labeled structured data, and the feature importance value is used to indicate the degree of influence of the feature on the result identified in the label; Normalize the feature importance value to obtain the feature importance vector; pass the feature importance vector to the sparse autoencoder network to affect the neuron weight; input unlabeled structured data into the sparse autoencoder network, to perform dimensionality reduction on unlabeled structured data. Through this application, it solves the technical problem that it is difficult to obtain labeled data in related technologies, and none of the existing solutions for extracting effective information from unlabeled data to improve the model prediction performance of the scene cannot meet the current needs.

Description

technical field [0001] The present application relates to the field of feature extraction, in particular, to a method and device for extracting dimensionality reduction features. Background technique [0002] In machine learning modeling scenarios, there will be a shortage of labeled data but a large amount of unlabeled data. In this case, if only a small amount of labeled data is used for modeling, good prediction results are often not obtained. For example, in the field of financial consumer credit, cashing out of orders often occurs, but the acquisition of such tag data requires relatively high costs in terms of time and labor costs. [0003] The current method to deal with this small number of samples is to use sample generation methods to increase the sample size, but this method is prone to over-fitting in the modeling process. [0004] In the case that the above-mentioned labeled data is difficult to obtain, effective information can be extracted from unlabeled data...

AI Technical Summary

Problems solved by technology

[0008] This application provides a method and device for extracting dimensionality reduction features to solve the problem that labeled data is difficult to obtain in related technologies, and none of the existing solutions for extracting effective information from unlabeled data to improve the model prediction performance of the scene Technical issues that meet today's needs

Images