Mobile terminal replacement prediction method, device and equipment and readable storage medium

Abstract

The invention discloses a mobile terminal replacement prediction method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining business data, constructing RandomForst and XGBOOST models, converting leaf node positions of a sample in the models into one-hot coding features, and generating a first feature; performing feature combination on the first features to generate second features, and combining the first features and the second features; selecting the merged features by adopting an integrated filtering method, and removing the features with poorcharacterization capability; constructing a sample weight function, and optimizing a model loss function; and constructing a logistic regression random forest based on a bagging framework, and performing mobile terminal replacement prediction through a model. According to the technical scheme, the method has the advantages of being high in representation capacity, low in cost, accurate in prediction and high in robustness, operators can conveniently and accurately market mobile phones to users, and high commercial value is achieved.

Description

technical field [0001] The present invention relates to the field of big data technology, and in particular to a mobile terminal replacement prediction method, device, equipment and readable storage medium. Background technique [0002] With the development of my country's mobile network and the popularization of 4G and 5G technologies, the smartphone market is developing rapidly. The popularity of smart phones has become unstoppable, and mobile phones are playing an increasingly important role in people's daily life. Communication, consumption, learning and other fields are inseparable from mobile phones, and mobile phones have become a necessity of life. The market share of mobile terminal replacements has become huge, and the frequency of user replacements has increased significantly. How operators use their massive user business data to accurately market mobile phones to users has become a strategic center for operators to expand their markets. [0003] In the existing ...

AI Technical Summary

Problems solved by technology

[0004] However, in the construction process of the existing replacement model, there are still many problems

First, using machine learning to build a model requires manual construction of features with strong representation capabilities, but different business scenarios require the construction of different artificial features, and the labor cost is too high

Second, the artificial features constructed may not be effective in the end. In the case of a large amount of data, it will take more time to verify the validity of the features.

Third, in the replacement data, the number of users who replaced their mobile phones is far less than the number of non-replacement users. The distribution of sample categories is unbalanced, and the sample weights are consistent by default. If the default optimization loss function is used, the model will be reduced. Generalization Ability in Minority Classes

Images