Travel network cell division method based on Simhash algorithm

Abstract

The invention relates to a travel network cell division method based on a Simhash algorithm. According to the method, the Simhash algorithm is utilized for processing texts and calculating the semantic fingerprints of the texts, the Hamming distance is used for comparing the distance between the semantic fingerprints of the texts, the text similarity is calculated, and therefore the purpose of clustering similar users is achieved. Dimensionality reduction is carried out on short-text high-dimensionality feature vectors, and the storage space occupied by the feature vectors is greatly reduced; meanwhile, the algorithm is easy and convenient to implement, short in calculation time and high in text processing speed, so that the division efficiency of complex network cells is improved, and the method has the great significance in predicting travel activity trend and development, providing travel service information, recommending travel routes and even predicting the travel peak.

Description

technical field [0001] The invention belongs to the field of data mining, in particular to a Simhash deduplication algorithm used as a clustering algorithm for community division in complex tourism networks. Background technique [0002] In recent years, complex networks have become a research hotspot in information science, sociology, physics, and even life science. Many systems in nature can be expressed in the form of complex networks, such as social networks, communication networks, and the Internet. Forums, BBS, Weibo, travel websites and other social platforms are widely used by travel enthusiasts because of their fast speed, low cost, and convenient use. Therefore, using these social platforms to communicate has gradually formed a complex tourism network. [0003] At present, there are different kinds of algorithms for complex network community division, one is the strategy of subgraph, such as spectral dichotomy, K-L algorithm; their disadvantage is that the size of...

AI Technical Summary

Problems solved by technology

The idea of ​​G-N algorithm is simple, but this method has a large amount of calculation, and the calculation time complexity is relatively high

The typical clustering methods are K-Means and K-Medoids. Although these two algorithms run fast, their disadvantages are that the selection of the initial center cluster will affect the clustering results, and they are easily affected by outliers, etc.

Representative algorithms for hierarchical clustering include BIRCH, CURE, and Chameleon. Its disadvantage is that it cannot be corrected once the process ends.

The disadvantage of grid clustering is that it is not very handy for processing large-scale data

Therefore, the current network community division methods all have different defects, so that their application is limited

Images