Research And Application Of Pollination Strategy In Clustering Algorithm

With the development of the Internet and the advent of the information age,a large amount of data has been generated in all walks of life.How to dig out useful information for people from massive data becomes particularly important.Cluster analysis technology is an unsupervised learning method and is also one of the commonly used analysis methods in data mining.The existing swarm intelligence adaptive clustering algorithm has the disadvantages of weak optimization ability,cluster intensiveness,low cluster comprehensive quality as well as poor universality.In view of the above problems,this thesis proposes an adaptive clustering algorithm which based on pollination strategy.In addition,in view of the shortcomings of traditional location selection algorithms that need to specify the number of clusters and the sensitive parameters,a method and system for merchant location selection which based on pollination-inspired clustering is proposed.The adaptive clustering algorithm based on pollination strategy is a biological heuristic clustering algorithm inspired by the interaction process of pollinating insects and plants in nature.This algorithm is different from the previous swarm intelligence algorithm.It not only avoids the waste of computing resources,but also improves the operating efficiency.In addition,two location update strategies are proposed to quickly update the data points to a more adaptable location.on the one hand,it enhances the algorithm's ability to optimize;on the other hand,it improves the clustering density and the overall quality of clustering.Among them,in the local position update strategy,the thermal kernel function in Laplace feature mapping algorithm and the application of adjacent weights are introduced as well as combined with the mean shift algorithm,it continuously updated the position of the data point in the two-dimensional grid.It also enhanced the adaptability of the algorithm,which making it suitable for multiple data sets.The test data set verifies the effectiveness of the proposed algorithm.In addition,the experimental results on various data sets show that the algorithm as a whole has better stability,convergence speed as well as clustering quality than other swarm intelligence algorithms.Moreover,the sensitivity analysis of each parameter of the algorithm is carried out in this thesis,which makes the parameter selection more reasonable.The adaptive clustering algorithm proposed in this thesis is simulated on the Geolife data set based on speed and pruning,and the algorithm is also applicable to the mobile terminal data set.In addition,it compared with other location algorithms,the clustering algorithm proposed in this thesis has the advantages of self-adaptation and insensitivity to noise.Finally,according to the experimental results and the geographic information system to analyze the user's lifestyle,consumption level and other factors to solve the problem of merchant's site selection.