Training method and device for classification model

Abstract

The invention discloses a training method and a training device for a classification model, which are used for improving data analysis efficiency. The training method of the classification model comprises the steps of: receiving sample data used for training the classification model, wherein the sample data comprises a plurality of sample features; determining a target feature subset from the sample data, and determining a high-dimensional sparse feature of the target feature subset by utilizing a high-dimensional sparse conversion method; determining target data complexity corresponding to the high-dimensional sparse feature of the target feature subset, wherein the target data complexity comprises a plurality of dimensions used for representing data features; determining a target classification algorithm corresponding to the target data complexity according to an established mapping relation between the data complexity and the classification algorithm, and determining target parameters corresponding to the target data complexity according to an established mapping relation between the data complexity and a hyper-parameter set of the target classification algorithm; and training the target classification algorithm according to the determined target parameters and the high-dimensional sparse feature of the target feature subset, so as to obtain the classification model.

Description

technical field [0001] The present application relates to the field of data processing, in particular to a training method and device for a classification model. Background technique [0002] With the advent of the big data era, information data is expanding day by day, and the market demand for efficient, robust and accurate analysis of massive data continues to expand. Such as off-grid prediction in the telecommunications field, medical diagnosis, credit rating in credit card systems, image pattern recognition, and network data classification. In this context, machine learning has been widely used, especially the classification method in machine learning is the most widely used. [0003] However, the use of classification methods faces many difficulties, among which feature selection, feature transformation, model selection and parameter tuning are the most difficult. Repeated attempts, modifications, and iterations are required, which makes the data analysis cycle long a...

AI Technical Summary

Problems solved by technology

[0004] However, because of this, there are many factors that affect data analysis, and the cost of positioning and debugging data analysis results is very high. Especially in big data scenarios, it takes a lot of time to calculate each data analysis, resulting in the entire data analysis cycle. Long, inefficient data analysis

Images