基于PCA与SVM的地力评价研究

发布时间：2018-06-03 07:10

本文选题：地力评价 + 主成份分析　；参考：《中南林业科技大学》2015年硕士论文

【摘要】：地力评价是对一定地域内土壤因素的综合评价,也叫做土壤肥力评价。目前,地力评价的方法有很多,主要包括层次分析法、经验判断指数法、灰色关联度分析法和模糊综合评价法等多种,计算机和信息技术的发展使得神经网络与SVM(支持向量机)成为地力评价的热点。本文首先介绍了国内外常用的地力等级评价方法,并分析了现有方法的不足,结合地力数据的特点,提出了基于PCA(主成份分析)与SVM(支持向量机)相结合的分类评价方法；然后简要论述了主成份分析和支持向量机的理论基础；运用隶属度函数计算出各评价指标的权重,克服了传统的主观设置权重的方法；求出各样本的地力等级,并作为评价模型中样本的标签；通过构建PCA-SVM等级评价模型,将PCA-SVM评价模型应用于地力等级评价实验中,并与单一的SVM评价模型、BP神经网络评价模型、Apriori半监督机器学习评价方法进行比较,根据实验结果分析该模型的可行性。实验证明,该评价模型不仅能够对地力相关数据进行约减,减少数据冗余,降低数据复杂度,而且在分类精度和稳定性等方面表现较佳,且减少了人为主观因素的影响,为地力等级评价提供了新的选择。
[Abstract]:Soil fertility evaluation is a comprehensive evaluation of soil factors in a certain area, also known as soil fertility evaluation. At present, there are many methods for the evaluation of earth force, including Analytic hierarchy process (AHP), empirical judgment Index (EPI), Grey Relational degree Analysis and Fuzzy Comprehensive Evaluation, etc. With the development of computer and information technology, neural network and SVM (support vector machine) become the focus of earth force evaluation. Firstly, this paper introduces the common methods of land force grade evaluation at home and abroad, and analyzes the shortcomings of the existing methods. Combined with the characteristics of soil force data, a classification evaluation method based on PCA (Principal component Analysis) and SVM (support Vector Machine) is put forward. Then it briefly discusses the theoretical basis of principal component analysis and support vector machine, calculates the weight of each evaluation index by using membership function, overcomes the traditional method of setting weight subjectively, and calculates the soil force grade of each sample. It is used as the label of the sample in the evaluation model, and the PCA-SVM evaluation model is applied to the evaluation experiment of the earth force grade by constructing the evaluation model of the PCA-SVM grade. The feasibility of the model is analyzed according to the experimental results compared with that of the single SVM evaluation model, the BP neural network evaluation model and the apriori semi-supervised machine learning evaluation method. The experimental results show that the model can not only reduce the data redundancy and complexity, but also improve the classification accuracy and stability, and reduce the influence of human subjective factors. It provides a new choice for the evaluation of soil fertility grade.
【学位授予单位】：中南林业科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S158;TP181

【参考文献】