Clustering method based on mathematical expression of som clustering model

Abstract

The invention provides a clustering method based on the mathematical expression of the SOM clustering model. This method analyzes the mathematical expression in MathML format, processes the mathematical expression into an expression tree form, and obtains the hierarchical position and sub-node information of each element of the expression; The elements are assigned weights; then the SOM neural network is used to train the mathematical expression sample set to obtain the corresponding SOM clustering model; finally, the SOM clustering model is used to cluster the mathematical expressions. The present invention uses the SOM neural network in the model-based method to cluster mathematical expressions. The network simulates the self-organizing feature mapping function of the human brain. The network structure is simple, and the features of the expressions can be represented by weights, which can be well integrated with The neural network input neuron fits, and using it as input data can get a good clustering effect.

Description

technical field [0001] The invention relates to the field of information retrieval, in particular to a clustering method based on the mathematical expression of the SOM clustering model. Background technique [0002] With the advent of the era of big data, a large number of scientific and technological documents emerge, and clustering, as an effective way to quickly obtain useful information from a large number of documents, makes it widely used in document information processing. Laith Mohammad Abualigah et al. started from the characteristics of the text, and used the feature selection method of the particle swarm optimization algorithm to cluster the text documents, which improved the clustering performance. Jiaming Xu et al. applied the self-learning convolutional neural network to text clustering, flexibly combined more useful semantic features, and combined with K-means clustering to obtain the best clustering results. K-means clustering is also applicable. By increas...

AI Technical Summary

Problems solved by technology

[0003] Common clustering methods include: Partitioning Methods represented by the K-Means algorithm, which need to determine the number of clusters and the initial clustering center before clustering; Hierarchical Methods require continuous Calculate the distance between samples and then merge them. Although the number of clusters does not need to be preset, the calculation complexity is high and the samples may be clustered into chains; density-based methods take DBSCAN as an example , you can find outliers in clustering, but samples with uneven density have a great influence on the clustering results, and a large number of samples will make the clustering convergence time longer; the grid-based method is not affected by The impact of the number of data objects, but it is more sensitive to parameters and prone to dimensionality disaster; the neural network based on the SOM (Self-Organizing Feature Maps) algorithm is a competitive learning unsupervised neural network, which will The relationship is imposed on the cluster centroid, and the clusters that are neighbors are more related to each other than the clusters that are not neighbors. This connection is conducive to the interpretation and visualization of clustering results.

Images