A method and device for automatic feature construction of structured data

Abstract

The invention discloses an automatic feature construction method and device for structured data. The method includes: S1: performing data processing on original data, the data processing at least including preprocessing, and the preprocessing at least including missing value processing; S2: after combining processing data, initialize the pre-built construction tree; S3: combine the processed data, perform a generation operation on the initialized construction tree, and obtain the feature generation method; S4: use the feature generation method to perform feature extraction on the preprocessed data, and obtain Feature generation results. The method and device for automatic feature construction of structured data provided by the embodiments of the present invention do not require training based on external data or specific domain knowledge, have a wide range of applications, low complexity, and can perform large-scale operations to cope with the huge search for automatic feature construction. Space; using a resampling strategy can improve the stability of the construction results and reduce the impact of noise.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to an automatic feature construction method and device for structured data. Background technique [0002] Feature engineering refers to the process of using domain knowledge to process the original data set so that machine learning can achieve its purpose (that is, the process of converting the original data into the training data of the model). Its purpose is to obtain better training data features. Data is the carrier of information, and the original data set may contain a lot of noise, resulting in unsuccinct information expression and unobvious hidden information, which is difficult to be captured by machine learning algorithms. Through a series of engineering activities, feature engineering uses more efficient features to represent this information, which not only preserves and strengthens the rules of the original data set, but also reduces the uncertainties, noise, an...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problems of the prior art, the embodiment of the present invention provides an automatic feature construction method and device for structured data, so as to overcome the need for model training based on external data or specific domain knowledge in the prior art, and the scope of application is small , and more complex issues such as

Images