Improved Particle Swarm Algorithm And Its Application In Data Mining

With the rapid growth of data, it is difficult to find required valuable information andtherefore data mining technology comes into being. Clustering is one of the most importantdata mining techniques, which divides the data into different clusters according to theirintrinsic characteristics. The data in the same cluster have the maximum similarity and thedata in the different cluster have the maximum dissimilarity.K-means algorithm proposed by the MacQueen is a classical algorithm to solve theclustering problem, and has been widely used in the field of data mining and knowledgediscovery. However, there are two major disadvantages for K-means algorithm. One is theresults of K-means clustering depending on the selection of the initial value, and the other isthat K-means algorithm based on gradient descent search is often trapped into a localoptimum. A clustering algorithm based on cooperation of particle swarm optimizer (PSO) andsimulated annealing(SA) is proposed in this paper, and is used in two different applications.The main work includes as follows:(1) To deal with the disadvantages of K-means and PSO, a clustering algorithm named asPSK-means based on cooperation of particle swarm optimizer (PSO) and simulatedannealing(SA) and probabilistic jumping property of SA is proposed. PSK-means canovercome the disadvantages of PSO as falling into local optimum and optimize the optimalsolution of particles and centroids to achieve optimal results. Simulation experiment is carriedout to validate good global convergence of PSK-means algorithm.(2) PSK-means algorithm is used in teaching quality evaluation system of computerexperiments. In the system, teaching evaluation scores can be collected from teachers andstudents and the procedure is conducted according to the evaluation indices. In the case ofevaluation of students and for the prepared evaluation scores, PSK-means algorithm is usedfor clustering analysis. A synthetical, accurate and objective evaluation is attained for students.Therefore, teachers will be clearly aware of students’ current situation and deficiencies, andthen take targeted measures to guide their learning. This can save time for users and improvethe instructional effects and learning performance.(3) PSK-means algorithm is used in analysis of traditional Chinese medical cases. Thereare a wide variety of traditional Chinese medical cases in Shandong traditional Chinesemedical cases system, including coronary heart disease, high blood pressure and otherdiseases. Firstly, the medical cases are selectively extracted according to type of illness.Secondly, PSK-means clustering is used on the prescription data of same disease to classify the prescriptions into different syndromes classes. At last, Apriori algorithm is used to mineassociation rule, find out different core prescriptions of types and provide a reference for thestudy of young traditional Chinese medical doctors.