Title :
A Method of Collapsibility Classification Based on Probabilistic Neural Network
Author_Institution :
Coll. of Inf. Commun. Eng., Changchun Univ. of Technol., Changchun, China
Abstract :
A model of probabilistic neural network (PNN) to classify the loess according to its collapsibility is suggested in this paper, in which five physical property indexes such as water content, dry density, void ratio, saturation degree and plastic liquid are taken as input neural cells and the output neural cell is coefficient of collapsibility. 76 groups of sample to be trained under different smoothing parameter as so to get an optimal parameter and obtained a suitable network with a good performance. The prediction result from PNN after training has a higher accuracy with 90% in 20 group testing samples, which is better than the accuracy predicted by RBF model. The result shows that the PNN is an effective classification method for collapsibility of loess and it is likely to implement in practice.
Keywords :
pattern classification; probability; radial basis function networks; PNN; RBF model; collapsibility classification method; dry density; input neural cells; optimal parameter; output neural cell; physical property indexes; plastic liquid; probabilistic neural network; saturation degree; void ratio; water content; Indexes; Neural networks; Probabilistic logic; Smoothing methods; Soil; Support vector machine classification; Training; RBF; collapsibility classification; probabilistic neural network;
Conference_Titel :
Computer Science & Service System (CSSS), 2012 International Conference on
Conference_Location :
Nanjing
Print_ISBN :
978-1-4673-0721-5
DOI :
10.1109/CSSS.2012.84
