基于机器学习—线性回归算法的收入与用户预测模型在审计项目中的应用  

Application of the revenue and user prediction model based on machine learning-linear regression algorithm in audit projects

作　　者：杨思祺 李淑兰 Yang Siqi;Li Shulan(Jingdezhen University,Jingdezhen 333400,China)

机构地区：[1]景德镇学院,江西景德镇333400

出　　处：《无线互联科技》2023年第23期94-98,共5页Wireless Internet Technology

基　　金：江西省教育厅科技项目,项目编号:GJJ2202418;景德镇市科技局项目,项目编号:20212GYZD011-03。

摘　　要：文章利用先进的大数据技术——线性回归算法构建主营收入与新增用户模型,研究运营商经营管理面临的各种风险。该模型选取了在网用户数、出账用户数、新增用户数、主营业务收入、预算收入、出账收入、欠费金额、APPU值等特征指标。模型经过多次迭代优化,按照最佳模型进行验证,通过模型预测值和实际数据进行比对验证。验证结果表明,年底累计收入和累计新增用户可以基于现有数据进行线性拟合,最终预测误差率保持5%之内。该模型可在财务收入计列合规性核查和年初收入与用户预算合理性2个场景中充分应用。Based on big data,this study constructed a relational model of the main business income and the new users with the application of linear regression algorithm,study the various risks faced by operators in operation and management.Which selected the number of online users,number of the users logout,number of new users,the main business income,the budget income,the income paid out,the amount of arrears,and the APPU value etc.for the index system of its operation.The model was optimized several times,verified according to the best model,and verified by comparing the predicted value of the model with the actual data.The research result shows that the cumulative revenue and cumulative new users at the end of the year can be linearly fitted based on the existing data,with an error rate less than 5%,this model can be safely used in verifying the financial income accounting compliance and reasonableness of the revenue and user budget at the beginning of the year.

关 键 词：线性回归 大数据 财务收入 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]