摘 要
随着大数据技术的快速发展，数据库中数据量急剧增长，传统分类算法在处理高维、海量数据时面临效率低、准确率下降等问题，亟需针对数据库特性进行优化。本文旨在研究面向数据库的分类算法优化方法，提升分类性能与计算效率。通过分析典型分类算法在数据库环境中的执行瓶颈，提出一种基于特征选择与索引优化的混合式分类框架，并引入数据库内置函数与并行计算策略以加速训练与预测过程。实验基于多个公开数据集，在关系型数据库系统中构建测试环境，对比多种主流分类算法的优化前后表现。结果表明，所提方法在保持较高分类准确率的同时显著提升了执行效率，尤其在大规模数据场景下具有明显优势。本研究的主要创新在于将数据库优化技术与机器学习算法紧密结合，提出了可落地的协同优化方案，为数据库内机器学习的工程实践提供了新的思路与实证支持。关键词：数据库内机器学习；特征选择与索引优化；混合式分类框架；并行计算策略；分类算法性能提升
ABSTRACT
With the rapid development of big data technologies, the volume of data in databases has increased dramatically, posing significant challenges to traditional classification algorithms in terms of efficiency and accuracy when handling high-dimensional and large-scale datasets, thus necessitating optimizations tailored to database characteristics. This paper aims to investigate optimization methods for classification algorithms oriented toward databases, with the goal of enhancing both classification performance and computational efficiency. By analyzing the execution bottlenecks of typical classification algorithms within a database environment, this study proposes a hybrid classification fr amework based on feature selection and index optimization, and further introduces built-in database functions and parallel computing strategies to accelerate the training and prediction processes. Experiments are conducted on multiple publicly available datasets within a relational database system to establish a testing environment, comparing the performance of various mainstream classification algorithms before and after optimization. The results demonstrate that the proposed method significantly improves execution efficiency while maintaining a high level of classification accuracy, particularly exhibiting notable advantages in large-scale data scenarios. The primary innovation of this research lies in the close integration of database optimization techniques with machine learning algorithms, offering a practical collaborative optimization scheme and providing new insights and empirical support for in-database machine learning engineering practices.
Keywords: In-Database Machine Learning; Feature Selection And Index Optimization; Hybrid Classification fr amework; Parallel Computing Strategy; Classification Algorithm Performance Improvement
目 录
摘 要 I
ABSTRACT II
绪 论 1
第一章 数据库分类算法优化的理论基础 2
1.1 分类算法在数据库中的应用需求 2
1.2 常见分类算法及其性能瓶颈分析 2
1.3 面向数据库优化的算法改进方向 3
第二章 数据库特征处理与优化策略 4
2.1 数据库特征选择方法比较 4
2.2 特征归约对分类效率的影响 4
2.3 数据预处理对模型性能的提升 5
第三章 分类算法在数据库中的并行化实现 6
3.1 并行计算框架的选择与适配 6
3.2 算法并行结构设计与优化 6
3.3 大规模数据下的实验验证 7
第四章 面向数据库的分类系统构建与评估 8
4.1 实验平台与数据集构建 8
4.2 优化算法的性能对比分析 8
4.3 系统稳定性与可扩展性测试 9
结 论 10
参考文献 11
致 谢 12
