Study On The Analysis Of Reader Behavior Using A Clustering Algorithm Based On Artificial Fish Swarm Algorithm

In the age of information, data storage capacity has increased day by day. In order to obtain valuableinformation from these erratic data seemingly, the technology of data mining was proposed. During thesuccessive century, the technology has achieved great development. Moreover, it has been extensivelyapplied in more and more fields. During this improvement procession, the study on clustering analysis hasbeen one of the major important tasks in the field of data mining.Emerging Swarm intelligence algorithm such as Particle Swarm Optimization, Bacterial ForagingOptimization in the field of optimization has made an enormous success in recent decades. Artificial fishswarm algorithm (AFSA) is one of swarm intelligence algorithm. Since it was proposed, the algorithm hasattracted the attention of experts and scholars. In order to improve its practicality, AFSA is widely appliedto different fields, including the field of data mining.The university library has the relatively stable readers, and a large sum of user records are generated inthe daily service. Library management system records are the behavioral outcomes that the readers to meetthe needs of individual, and also these are the powerful evidence of library resources recognized by readers.Some related knowledge with users is hidden in the reader information, book information and userborrowed records.Due to some commercial interests, the study on readers’of university behavior analysis is not enoughat present. By learning clustering analysis algorithm and artificial fish swarm algorithm, in this paper, weproposes a new hybrid algorithm based on improved AFSA and K-means for data clustering. We use thenew algorithm as the core. This paper also proposes three kinds of models of behaviors analysis of readers,which contribute to certain basis about the grasp of the rules of readers lending from library, the degree ofthe preference of library books, and the improvements of the services models.First, we design a new hybrid algorithm based on improved AFSA and K-means for data clustering.The algorithm is improved from the following several aspects. To improve the parameter of visual, thealgorithm takes the form of vector, and adopts the standard deviation of each attribute value for clusteringsample as the benchmark. It integrates particle swarm strategy into the follow operator to guide the learningof artificial fishes. In the iterative process, it randomly selects a small part of artificial fish which is not inthe optimal state, to execute the K-means behavior, according to nearest neighbor principle. Experimentalresults prove the effectiveness of our algorithm. This algorithm combines the advantages of artificial fishswarm algorithm and K-means algorithm, not only can overcome sensitive issue for selecting the initialcluster center about the K-means, but also improve the later rate of convergence of the AFSA, can obtainthe best quality in relatively shorter time.Second, we process the reader information, library book information and user borrowed record data.Its steps include sorting, cleaning and transformation. According to research needs, this paper puts forwardthree kinds of reader behavior analysis model: reader borrowed volume model, book borrowed volumemodel and single borrowed characteristics model.Finally, applying this improved algorithm into these three analysis models to discover the insideregular pattern characteristics, to instruct the readers about the behaviors of the lending, to improve the useratio of the resources, optimizing the models of services and so on, to provide the fact basis.