基于机器学习的不确定数据增量式挖掘算法  

Incremental Mining Algorithm for Uncertain Data Based on Machine Learning

作　　者：文明瑶[1] 廖伟国 WEN Ming-yao;LIAO Wei-guo(Zhujiang College,South China Agricultural University,Guangdong Guangzhou 510900,China)

机构地区：[1]华南农业大学珠江学院,广东广州510900

出　　处：《计算机仿真》2021年第11期290-294,共5页Computer Simulation

基　　金：2017年广东省高等教育教学研究和改革项目(706)。

摘　　要：以实现数据增量式精准挖掘为目的,提出基于机器学习的不确定数据增量式挖掘算法。以机器学习算法中的模糊c-均值聚类(FCM)算法为基础,通过主成分分析法筛选原始数据集中指标,利用Relief算法计算指标权重,实现FCM算法改进。改进FCM算法通过阈值定义目标函数,经样本数据分类、特征提取和聚类,使目标函数达到最小值,实现数据挖掘。实验结果表明,上述算法的数据样本分类符合率可达99.28%,分类准确率在98%左右,且分类耗时短、效率高;特征提取能力受数据量增加影响较小;在数据增量情况下,改进算法增量式挖掘准确率保持在95%~98%之间,且所需迭代次数少。In order to realize incremental accurate data mining, an incremental uncertain data mining algorithm based on machine learning is proposed. Based on the fuzzy c-means clustering(FCM) algorithm in machine learning algorithm, the indexes in the original data set were selected by principal component analysis, and the index weight was calculated by relief algorithm to improve the FCM algorithm. Then, the improved FCM algorithm defined the objective function through threshold, and made the objective function reach the minimum value through sample data classification, feature extraction and clustering to realize data mining. The experimental results show that the data sample coincidence rate of the algorithm can reach 99.28%, the classification accuracy is about 98%, and the classification time-consuming us short and the efficiency is high;The ability of feature extraction is less affected by the increase of data volume;In the case of data increment, the accuracy incremental mining of the improved algorithm remains between 95% ~ 98%, and the number of iterations is less.

关 键 词：机器学习 增量式 数据挖掘 聚类 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]