|
This paper introduces the concept of germplasm resources and research , the importance of protection , based on the analysis of the general concepts and methods of data mining and knowledge discovery , more comprehensive summary of the type of cluster analysis algorithms for data mining and principle . Selected on the basis of the theoretical research , analysis and comparison of the different types of clustering algorithms , k-means, k-mediods, FCM, CURE, SOM, GA involving division method , the fuzzy method , hierarchical method , artificial intelligence and in-depth study of the six kinds of machine learning methods clustering algorithm , system analysis and design , object-oriented programming method application C # programming language implementation , and ultimately the formation of a more general integration of a variety of methods of clustering analysis software system . The article also developed software on the validity of the test , the clustering validity evaluation function . By clustering of the two - dimensional random data points indicate that the test procedure can meet the multidimensional data clustering requirement under normal circumstances , each of the above clustering algorithm clustering is effective , SD cluster validity evaluation index is scientifically valid . The study found : SOM self-organizing neural network compared to other algorithms are more efficient , more stable results , to good effect , the great value of further in-depth study ; SD validity index not only can be used to evaluate the clustering effect , good or bad , and can be used in guiding the determination of the optimal number of clusters , with a further in-depth study of the value . Finally , this study advance research cluster analysis - based on the cluster analysis of data mining software system - Populus the 150 clones blades factor data , explore of the the germplasm resources Populus data mining method , preliminary complete data mining to extract the typical class knowledge \
|