基于流数据的单指标众数模型推理及实证研究  

Reasoning and Empirical Study of Single Indicator Mode Model Based on Stream Data

作　　者：朱桂玲 郭召 ZHU Guiling;GUO Zhao(Zhaotong College,Zhaotong University,Zhaotong Yunnan 657000;Zhengzhou Urban Construction Vocational College,Zhengzhou Henan 452100)

机构地区：[1]昭通学院,云南昭通657000 [2]郑州城建职业学院,河南郑州452100

出　　处：《中国科技纵横》2025年第3期171-173,共3页China Science & Technology Overview

基　　金：云南省教育厅科学研究基金项目“基于流数据的单指标众数模型的推理及实证研究”(2024J1076);昭通学院课程思政示范项目:《应用回归分析》;昭通学院应用型特色示范课程《应用回归分析》。

摘　　要：在哲学、经济学、教育学、农学、医学、艺术学、建筑工程以及军事科技等领域,高维数据大都存在倾斜。如果忽略数据的倾斜性而直接进行研究,就会导致结论失效。为了解决这个问题,需要使用单指标众数模型。在现实生活中,不仅包括高维数据,也包括大量、快速、连续到达的流数据。本研究拟探讨基于流数据的单指标众数模型推理及实证分析,其不仅是众数回归模型统计诊断的推广,也与模型中当前数据批次和历史数据批次的在线程序更新有着密切联系。在流数据环境下,本文使用逻辑回归模型的负对数似然函数对单指标众数模型的在线参数估计进行研究,提出该模型的在线LASSO估计算法。In fields such as philosophy,economics,education,agriculture,medicine,art,as well as construction engineering and military technology,most of these high-dimensional data are skewed.If the skewness of the data is ignored and studied directly,it will lead to invalid conclusions.To address this issue,a single index mode model is proposed.In real life,the data we face is not just high-dimensional data,but a large amount of rapidly and continuously arriving stream data.This study aims to explore the inference and empirical analysis of a single indicator mode model based on stream data.This problem is not only the extension of statistical diagnosis of mode regression models,but also closely related to online program updates of current and historical data batches in the model.To illustrate the proposed method,we use the negative logarithmic likelihood function of logistic regression models to study the online parameter estimation of single indicator mode models in a stream data environment,and propose an online LASSO estimation algorithm for this model.

关 键 词：流数据 在线参数估计 LASSO估计 单指标众数模型 

分 类 号：O212.1[理学—概率论与数理统计]