基于PCA降维的MNIST手写数字识别优化  

Optimization of MNIST Handwritten Digit Recognition Based on PCA Dimensionality Reduction

作　　者：田春婷 TIAN Chunting(School of Information Engineering,Lanzhou Petrochemical University of Vocational Technology,Lanzhou 730207,China)

机构地区：[1]兰州石化职业技术大学信息工程学院,甘肃兰州730207

出　　处：《现代信息科技》2024年第16期64-68,共5页Modern Information Technology

基　　金：甘肃省教育厅高校教师创新项目(2023A-205)。

摘　　要：PCA数据降维技术广泛应用于数据降维和数据的特征提取,可以很大程度上降低算法的计算复杂度,提升程序运行效率。文章将MNIST原始数据集和对原始数据集进行PCA降维处理之后的数据集作为样本,分别采用K-邻近算法、决策树ID3算法、SVC分类模型,以及选取不同分类算法作为基础分类器的集成学习方法,实现手写数字识别。在对MNIST数据集进行PCA降维前后,以及不同分类算法和模型执行结果的时间复杂度与预测准确率进行比对与分析,进一步强化与优化手写数字识别准确率等各项指标。PCA data dimensionality reduction technology is widely used in data dimensionality reduction and feature extraction,which can greatly reduce the computational complexity of algorithms and improve program efficiency.This paper takes the MNIST original dataset and the dataset after PCA dimensionality reduction as samples,and uses K-Nearest Neighbor algorithm,Decision Tree ID3 algorithm,SVC classification model,as well as Ensemble Learning methods that select different classification algorithms as basic classifiers to achieve handwritten digit recognition.It compares and analyzes the time complexity and prediction accuracy of different classification algorithms and models before and after PCA dimensionality reduction on the MNIST dataset,further enhances and optimizes various indicators such as handwritten digit recognition accuracy.

关 键 词：PCA降维 MNIST手写数字识别 K-邻近算法 决策树 SVC分类模型 集成学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP181[自动化与计算机技术—计算机科学与技术]