Structured data general modeling method based on Transform

• Main Authors: GUO YING, LI XIWU, SUN GUANGYUAN, LIANG RONGHUA, LIANG HAORAN, XIONG YUANYUAN, DIAO KEHONG
• Format: Patent
• Language: Chinese; English
• Published: 21.07.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

A structured data general modeling method based on Transform comprises the steps that firstly, irrelevant features of original data are removed, then different embedding methods are used for category features and numerical value features, feature vectors obtained after the category features and the numerical value features are embedded are spliced, and then a model is obtained; the method comprises the following steps: splicing feature vectors of a data set, inputting the spliced feature vectors into a Transform + neural network (an improved converter Transform) and an MLP + neural network, according to the Transform + neural network, adding a Leaky Gate in front of the original converter Transform and adding the MLP + neural network behind the original converter Transform, finally distributing different weights for output values of two modules, and meanwhile, in order to cope with the problem of class imbalance of the data set, determining whether the data set is a data set or not. According to the method, a