Application of factor analysis and probabilistic neural network model on evaluation of the slope stability

Slope stability analysis is a complex problem concerning system engineering, and slope stability evaluation directly affects the safety and economical efficiency of slope engineering. In order to realize rapid, efficient and accurate evaluation of slope stability, multiple evaluation indexes should be considered which exist more or less correlations which lead to the overlapping of parameter information. An improved factor analysis method is proposed to reduce the dimensionality of the slope stability index data. Three new indexes are extracted to conduct an overall evaluation of slope stability. The indexes undergoing factor analysis are independent of each other and can meet the requirement of adopting the Gaussian function as the radial basis function in the sample layer of PNN. On the basis of factor analysis, the PNN model for slope stability evaluation is established, which is applied to 39 typical slope stability evaluations. The predicting results show that the PNN model still presents a favorable predictive effect under 5 different training and test samples, and the correct judging ratio is 100%, 94.87%, 94.87%, 84.62% and 84.62%, respectively, which verify the evaluation results of factor analysis on slope stability and indicate that the combination of factor analysis and PNN model can provide a good thinking for slope stability evaluation in geotechnical engineering.