基于机器学习的猪场用户流失分析

Churn analysis of pig farm users based on machine learning

关键词:  猪场  互联网+农业  用户流失  机器学习  决策树

Key words:  pig farm  internet+agriculture  users churn  machine learning  decision tree

摘要:

为实现养殖互联网平台用户的自动化运营，防止用户流失，采用区间估计和机器学习建模方法，对猪场用户的平台使用情况进行用户流失分析。结果表明：1）猪场规模越小，用户流失的概率越大，其中500<母猪数x≤1 000的猪场规模用户流失的概率最小，属于忠诚用户；2）使用决策树算法、kNN算法、贝叶斯分类算法进行猪场用户流失分析建模，3种分类算法中，从平均识别率（F₁度量）考虑，决策树模型的平均识别率（0.93）高于kNN模型（0.91）和贝叶斯分类模型（0.80），选择决策树算法用于平台用户流失分析建模是可行的。本研究所构建的猪场用户流失分析模型，可为养殖互联网平台的猪场用户研究提供可靠的数据模型，实现平台用户的自动化运营，也可为其他类型的用户研究和产品设计提供参考。

Abstract:

In order to realize the automatic operation of users and prevent user churn,interval estimation and machine learning modeling methods are adopted to analyze user churn by data records of pig farm users.The results show that:1) The smaller the pig farm is,the greater the possibility of user churn is.The least possible churn of pig farm users,whose number of sows x is at 500<x ≤ 1 000,belong to loyal users;2) Comparing three different classification algorithms,e.g.the decision tree algorithm,kNN algorithm,Bayesian classification algorithm in pig farm users churn modeling,the average recognition rate of decision tree model (0.93) is higher than the kNN model (0.91) and Bayesian classification model (0.80).Therefore,choose the decision tree algorithm for platform user churn analysis is feasible.The research indicates that the churn analysis of pig farm users model could provide customized services for users of different sizes,provide a reliable data model for user research of other categories and realize the automatic operation of platform users.