High-order feature automatic generation method, system and device and medium
An automatic generation, high-level technology, applied in computing models, machine learning, computing, etc., can solve problems such as opacity and non-interpretation of human beings
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0065] like figure 1 As shown, this embodiment discloses a method for automatically generating high-order features, including the following steps:
[0066] Step S101. Obtain an input feature set, which contains several input features; each of the input features has a corresponding name and meaning;
[0067] Step S102, generating high-order features by performing operator operations on the input features in the current input feature set;
[0068] The operator in this embodiment includes the name of the operator, the meaning of the operator, and the execution mode of the operator, and a cross feature candidate is generated according to the operator. Operators mainly include unary operators, binary operators and multivariate operators.
[0069] Step S103, adding each generated high-order feature to the input feature set to form several sets of candidate feature sets, and evaluating the several sets of candidate feature sets using the selected machine model;
[0070] Step S104,...
example 1
[0079] Example 1: For example, the newly generated feature name is [age][disc5]. Age is the parent feature, representing the age of the user. Look up the table, disc5 is a unary operator, which means feature discretization, then the new feature is expressed as discretizing age into 5 levels.
example 2
[0080] Example 2: The newly generated feature name is [[age][disc5], gender, level][groupThenAvg]. Among them, [age][disc5], gender, and level are the parent features, and [age][disc5] is a new feature after the unary operator in (1), and it also acts as the parent feature in the process of feature generation. Features, gender indicates the user's gender, and level indicates the user's consumption level. Look up the table, groupThenAvg is a multivariate operator, and the meaning of the feature is to take the average after the feature is grouped, then the new feature is expressed as grouping by the two features of [age][disc5] and gender, and then take the average of the level in the group, the feature describes The average consumption level of users in different ages and genders.
[0081] Use the model to evaluate candidate features. Since each candidate feature needs to be evaluated individually, a simple and fast model will greatly improve time efficiency. Logistic regressi...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More