Prediction of Loess Collapsibility Coefficients Based on PCA and ANFIS
Ning An · Journal of Shenzhen Polytechnic · 2011
Guided by artificial intelligence theory,a prediction method of loess collapsibility coefficients was proposed based on principal component analysis(PCA) and adaptive neuro-fuzzy inference system (ANFIS).The principal component of physical indicators of the loess was extracted based o an analysis of principal components to eliminate the correlation between variables and reduce the amount of input.An ANFIS model was then established using the high adaptability of neural network and the reasoning faculties of fuzzy inference system,offering a new prediction method of collapsible loess was suggested.By comparison of the measured data and the predicted data,the average error was 0.29%and the maximum error was 20%,both of which can be accepted in actual projects.Its applications also indicated that this prediction method was feasible.