Multilevel index projection pursuit dynamic clustering method and device

Abstract

The invention relates to a multilevel index projection pursuit dynamic clustering method and device. The method includes the steps: building a multilevel evaluation index system according to difference of domains and targets; determining the weight of indexes in the multilevel evaluation index system; processing noise in real-time tuple data; building a projection pursuit clustering model based on the processed real-time tuple data and the weight of the indexes; realizing projection pursuit dynamic clustering for the processed real-time tuple data based on a genetic algorithm and the projection pursuit clustering model. The data can be rapidly, efficiently, objectively and dynamically clustered.

Description

Technical field [0001] The present disclosure relates to the technical field of high-dimensional data clustering, and in particular, to a projection pursuit dynamic clustering method and device for multi-level indicators. Background technique [0002] Traditional statistical models mostly use the confirmatory data analysis idea of ​​"hypothesis-simulation-prediction". This kind of analytical idea is difficult to adapt to the ever-changing objective world, and it is impossible to truly find the internal connections and laws of data. When it is used in high-dimensional, Non-linear, non-normally distributed data predictive modeling is more difficult to have good results. [0003] Agricultural data usually has the characteristics of large quantity, multi-level indicators, and multi-dimensionality. How to scientifically and effectively cluster and evaluate different data in the agricultural field has always been a key and difficult problem. Because these agricultural data are too huge,...

AI Technical Summary

Problems solved by technology

However, ordinary projection pursuit clustering methods cannot reflect the importance of actual index experience

At the same time, the traditional AHP (Analytic Hierarchy Process, Analytic Hierarchy Process) has the problem of too much subjectivity in evaluation, which may affect the evaluation and selection

In 1981, Friedman and Stuetzle proposed a multiple smoothing regression technique to realize PPR (Projection Pursuit Regression, projection pursuit regression), but this method is computationally complex and difficult to program, making it difficult to popularize and use

Images