Research Into Responsibility Problems With Elimination Of Poverty By Vocational Education In Rural Areas

As an unsupervised learning method, clustering analysis is one of the key research fields in data mining, and has been applied in a variety of areas in recent years. While rough set theory (RST) is a new mathematical tool for depicting incomplete and uncertain knowledge. It can also discover connotative knowledge, and reveal potential rules without any additional knowledge about the original data. The validity of RST has been proved in many applications of data mining. Due to these advantages, RST has been received more and more attentions from researchers. Some applied research is completed in the dissertation on the application of RST in the clustering analysis. The main contributions are as follows:1) Considering the problems in existing clustering algorithms, a new similarity measure is defined, based on which the discernibility ability is used to measure the importance of attributes, and thus a weighted rough clustering algorithm is proposed to deal with the categorical data. The theoretical analysis and experimental results show that the proposed clustering algorithm can deal with the categorical data, and does not need to be given the number of cluster, especially, it improves the cluster quality.2) Considering the limitations of classical clustering algorithms on high dimensional data, an approximation attribute reduction algorithm is proposed based on indiscernibility degree under similarity relation. Firstly, indiscernibility degree is used to evaluate the importance of attributes, and then the unimportant or irrelevant attributes are filtered, which improve the performance of clustering algorithm, and is very helpful for understanding the clustering result. At last, we test our algorithm versus PCA algorithm on UCI datasets, the experimental results show the proposed algorithm can select the important attributes without reducing the clustering accuracy.3) According to the case of soybean seed, growth situation, growth environment, regional disaster and so on, related to the use of the proposed algorithm, the growth trend of soybean were predicted classification, the result is correct. Study confirmed that the algorithm can be used to forecast trends in the growth of crops, especially soybean crops can be used for predictive analysis. The work of this dissertation extents the application fields of clustering analysis, and provide new methods and techniques for clustering analysis based on RST.