Study on System Inference of Probability Distribution of Geotechnical Parameters

With the help of data mining technology, the selection of probability distribution of geotechnical parameters for test samples was systematized. 19 groups of geotechnical parameters for test samples were selected, the beta distribution and the normal information diffusion method were chosen as two methods to infer the probability density of geotechnical parameters. Based on the fuzzy C-means clustering method in data mining technology, 19 groups of geotechnical parameters for test samples were clustered into 4 categories according to the numerical eigenvalues of samples, and then the inference method of probability density functions was determined according to the distribution law of each category for samples. For test samples with serious fluctuation and discreteness, an optimal determination method on window width was proposed, the method took the minimum mean square error as the constraint goal and changed the size of window width on the basis of window width. When the probability density function was deduced by the normal information diffusion method, in order to determine window width, an optimal determination method on window width of samples by changing the width value of the window on the basis of the minimum mean square error was proposed. For a group of new test samples, it was probable to infer the density function of samples by the numerical eigenvalue.