Study. Neuronal Morphology Classification Based Pca-clustering Algorithm

Neurons, in the brain, is the basic unit of the composition, understand the neuron all aspects of the factors, for example, morphological structure and function, to speed up the development of the brain have a vital role. In this paper from the structure of the form neurons, in the PCA algorithm and clustering algorithm based on the classification problem form neurons, makes a deep research. Establish the clustering algorithm PCA-clustering algorithm and PCA-clustering algorithm model for neurons shape classification problem solving, and on the three algorithm in the results made the corresponding comparisonPCA-clustering algorithm is improved in the PCA-clustering algorithm on the basis of the improvement, it inherits the clustering algorithm and the PCA-all the advantages of the clustering algorithm, slam the door clustering algorithm and the PCA-the deficiency of the clustering algorithm. At first, this paper try to use a conventional clustering algorithm neurons classification problem, it was found that the conventional clustering algorithms solve neurons classification problem there exists limitation, namely the characteristics of sample data correlation between variables if too high, data redundancy spend, for classification results will have a great effect on, in order to solve this problem, the PCA algorithm fusion with the clustering algorithm, PCA features a orthogonality, namely by PCA algorithm sample data for its main composition of correlation between no longer so, solve the sample data characteristic variables of the correlation between the impact. Secondly, in the PCA-clustering algorithm process, sometimes will meet the information provided to sample to bring better classification the effect, therefore, the author put forward the improvement, in order to "dig out the samples from more information (characteristics), introduce a entirely converted into a plane right Angle coordinate system method, the polygon conversion entirely. Finally using conventional clustering algorithm, PCA-clustering algorithm and PCA-clustering algorithm to improve the neurons sample data are calculated, and the accuracy of the way the contrast, found that conventional clustering algorithm for the classification of neurons effect is not ideal, the best conventional clustering algorithm for28.57%accuracy, PCA-clustering algorithm and PCA-clustering algorithm of improvement effect and correct rates were85.71%and87.76%respectively, PCA-clustering algorithm improved PCA-clustering algorithm has certain advantages.