Research On Improved Clustering Algorithms And Its Application In The Analysis Of Achievement

The data mining is called the backbone technology of the future information processing, it changes the way of the human using data by brand-new concept. The data mining is refers to the process of from the large amounts of data to extract useful information and knowledge. Here requests the data sources to be huge, real, includes the noise. The information and the knowledge which is discovered is latent and hides behind the mass data, is user interesting, apprehensible and exercisable knowledge. Cluster analysis is a basic assignment of data mining, Clustering is grouping the data object into many species or the cluster, divisiory principle is between the object of the same cluster has the high similarity, but the object of the different cluster has big difference.The subject of the thesis is based on the project"embedded intelligent online teaching platform"of Department of Information Industry of Liaoning Province, After analyzing and comparing the data mining technology deeply, An improved clustering approach algorithm is presented, combine this algorithm with the educational statistics to extract useful information from mass grade data. I develop analysis of achievement system of embedded intelligent online teaching platform, which realize automatic generation of the student achievement analysis and the examination paper quality rating report.The traditional K-means algorithm based on genetic algorithm can't get globally optimal value,because it is premature quickly in genetic process and all the population are stalling in the end of algorithm. An improved clustering approach based on genetic algorithm is presented. This algorithm applies an improved adaptive genetic algorithm based on simulated annealing algorithm makes the fitness properly, and then adaptively adjusts crossover and mutation probability and select with cross generational elitist selection, at the same time use the classical K-means algorithm, The optimal cluster centers can be searched by this algorithm.The paper uses Java language to carry on the experiment which compares with K-means algorithm and the traditional K-means algorithm based on genetic algorithm, Experimental results demonstrate that this algorithm effectively avoide easily falling into the local optimum because of the effect of the selection of the initial cluster center, and eliminate sensitivity to the data of isolated point ,at the same time overcome disadvantage of the traditional K-means algorithm based on genetic algorithm falling into the local optimum because of premature convergence. Theoretical analysis and experiment indicate that, this algorithm is better than K-means algorithm and the traditional K-means algorithm based on genetic algorithm. The improved K-means clustering algorithm based on genetic algorithm applied to Analysis of Achievement of examination papers's quality, and combining on-line teaching, thus discovers that the test question deficiency of embedded intelligent online teaching platform, accordingly further improves the quality of the examination papers ,thus achieves the satisfying intellectualized teaching effect.