Multi-model fusion method

A fusion method, multi-model technology, applied in character and pattern recognition, instruments, computer parts, etc., can solve the problems of low generalization ability, complex iteration, training not helpful, etc. Reduce the effect of iterative complexity and strong generalization ability

CN109948668AInactive Publication Date: 2019-06-28成都新希望金融信息有限公司
2 Cites 11 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2019-06-28
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a multi-model fusion method, belongs to the technical field of machine learning, and solves the problems of less dimension, low precision and complex iteration in traditional model classification. The method comprises: acquring and processing a data set containing user characteristics; based on an XGBoost algorithm and two different Y value intersection interval settings, selecting different derivation methods to derive characteristics in the processed data set; inputting the data set features and all the derived features into two XGBoost models as a new data set for training and optimization, and outputting the leaf node number of each training set sample at each tree by the two XGBoost models after optimization; Training One-hot coding according to leaf node number set, and enabling the leaf node number output by each XGBoost model to be converted by the trained One-hot code to obtain all LR features corresponding to the training set samples, and training an optimization LR model based on the LR features corresponding to all the training set samples; And obtaining an XGBoost + LR fusion model. The method is used for classifying users.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] A multi-model fusion method is used for classifying users and belongs to the technical field of machine learning. Background technique

[0002] Machine learning is a multi-field interdisciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. Specializes in the study of how computers simulate or implement human learning behaviors to acquire new knowledge or skills, and reorganize existing knowledge structures to continuously improve their performance.

[0003] Machine learning is a subset of artificial intelligence and the core of artificial intelligence. It is the fundamental way to make computers intelligent. Its application pervades all fields of artificial intelligence. It mainly uses induction and synthesis rather than deduction. At its core, machine learning is "using algorithms to parse data, learn from it, and then make a decision or prediction about ...

Examples

Embodiment

[0049] The XGBoost+LR fusion model processing flow includes two processes, each of which includes four processes including data preparation, feature engineering, model training, and result output, such as figure 1 As shown, for example: the data preparation (that is, data processing) step is mainly to clean and reshape the data set; the feature engineering is derived from the user's multi-dimensional features, and is selected through feature importance, and the top 100 features are respectively Pass in two XGBoost models, and after traversing 400 trees, record the number of leaf nodes of the training set samples in each tree (referring to the tuned XGBoost model, the number of training set samples in the leaf nodes of each tree), and pass One-hot encoding conversion to obtain all the LR features corresponding to the sample, then train and tune the LR model, and output the model result, which is the improved Gaussian transformation of the result. Specific steps are as follows:...