Landslide monitoring and analysis is the prerequisite and foundation of landslide prediction, and the results of landslide prediction are directly related to the distribution of monitoring points on landslide. In the landslide prediction, it is necessary to model the mathematical model based on the monitored data. It is a complicated and cumbersome to use the analysis of the points separately. Therefore, before the landslide monitoring point modeling, the fuzzy monitoring analysis of each monitoring point, mathematical modeling of each kind of monitoring points, and then use the built mathematical model for deformation analysis, which can reduce the follow-up calculation workload, so as to improve work efficiency. After fuzzy clustering, through the classification of deformation points, for understanding the landslide deformation of the partition has a very important reference value. The fuzzy clustering analysis procedure is as follows.

Establish the data matrix: set the domain for the classification of the object, each object by m indicators that its character, that,

,.

Thus, the original data matrix is

In practical problems, different data generally have different dimensions. In order to compare the amount of different dimensions, it is usually necessary to make appropriate changes to the data. There are generally translational standard deviation transform, translation range transformation and logarithmic transformation.

Set the domain to, , according to the traditional clustering method to determine the similarity coefficient, the establishment of fuzzy similarity matrix. The method of calculating the similarity degree of and mainly uses the similarity coefficient method, distance method and other methods of traditional clustering analysis. According to the characteristics of the problem, select a different formula to calculate.

According to the fuzzy matrix obtained by the calibration, only a fuzzy similarity matrix R is not necessarily transitive, that is, R is not necessarily a fuzzy equivalent matrix. In order to classify, it is also necessary to transform R into a fuzzy equivalence matrix. The quadratic method can be used to pass the closure, and the fuzzy similarity matrix is transformed into a fuzzy equivalent matrix.

Starting from the fuzzy similarity matrix R, the squares are obtained in turn, that is, , When the first occurrence of

(indicating is transitive), is the required transitive closure.

1) Data matrix

Using the landslide GPS monitoring data, see Table 1.

Get the original matrix of landslide monitoring data, see Table 2.

2) Data standardization

The data are converted using the translation range method.

3) After the data is standardized, see Table 3.

Calculate the degree of similarity, where the maximum and minimum methods in the similarity coefficient method are used.

The similarity matrix is calculated according

to Equation (3). indicates the degree of similarity between the i and j points.

Landslide monitoring point deformation fuzzy similarity matrix

The fuzzy equivalence matrix is calculated as:.

From large to small to cluster analysis.

When, U can be divided into five categories P1, P2, P3, P4, P5

When, U can be divided into {P1, P2}, {P3}, {P4}, {P5}

When, U can be divided into P1, P2, P3, P4, P5

When, U can be divided into {P1, P4}, P2, P3, P5

When, U can be divided into P1, {P2, P4}, P3, P5

When, U can be divided into {P1, P2, P3, P4}, P5

When, U can be divided into P1, P2, P3, P4, P5.

In the actual monitoring situation, the P1, P3 and P5 monitoring points are located at the front, middle and trailing edges of the landslide, while the P2 and P4 points are located on both sides of the central monitoring point P3. From the classification of results can be seen, when, points P2 and P4 for a class, the other monitoring points were a class.