Quantum clustering method based on nearest neighbor KNN and improved wave function

Abstract

The invention provides a quantum clustering method based on a nearest neighbor KNN and an improved wave function, and the method comprises the steps: obtaining original data of a group of to-be-classified sample points, carrying out the normalization of the original data, determining the input parameters of a quantum clustering model based on the nearest neighbor KNN, calculating the wave function parameters of all sample points, and carrying out the calculation of the wave function parameters of all sample points; the wave function parameters comprise the steps of calculating scale parameters and shape parameters of distribution obeyed by wave functions, calculating potential energy surfaces of quantum clustering, and determining the classification number and classification boundaries according to the calculated potential energy surfaces. The method provided by the invention inherits all advantages of a quantum clustering method, is more suitable for classifying data obeying Weibull distribution, provides a new choice for data classification, does not need to manually give any input parameter and does not need to give a classification label of sample data at the same time, and can be applied to the field of data classification. The input parameters of the quantum clustering model can be calculated, the practicability is high, and the accuracy is high.

Description

Technical field [0001] The present invention belongs to the field of clustering technology, relates to the classification of data, in particular a quantum clustering method based on the nearest neighbor KNN and improved wave function. Background [0002] Quantum clustering method is a kind of clustering model based on quantum mechanics, the core idea of the model method is that particles tend to be at the lowest potential energy value, and the same class of particles will gather near the same potential energy minimum point, so it is possible to calculate the potential energy surface of the particle is located for cluster analysis, each potential energy surface of the minimum value point represents the center point of a cluster, and the number of minimum points of the potential energy value represents the number of clusters, and finally through the potential energy value to determine the cluster center point to which the particle belongs. Compared with other clustering methods, qu...

AI Technical Summary

Problems solved by technology

Among them, the nearest neighbor KNN method can calculate the input parameters without providing sample labels, but its calculation still needs to artificially give the parameter of the number of neighbors, which makes it difficult for quantum clustering to become a parameterless clustering method in the true sense. ; Although the "pattern search" method no longer requires artificially given parameters, this method needs to provide sample classification labels, which is contrary to the core characteristics of quantum clustering as unsupervised learning

Images