Research On Adaptive Clustering Algorithm Based On Insect Pollination

Based on the characteristic data of the object itself,the adaptive clustering algorithm does not need to specify the number of clusters in advance,and through the evolutionary design of the mutual connection of the data objects,the algorithm can automatically find the number of clusters and the data contained in the cluster,and the adaptive clustering algorithm has become a hot topic in the field of data mining in recent years.For the current adaptive clustering algorithm based on swarm intelligence,the algorithm lacks the the mechanism of intelligent evaluation in the process of searching data objects,which results in wasting a certain amount of computing resource.In addition,it is difficult to enhance the fitness of data just using the random moving rules and so the quality of clustering result and efficiency of algorithm both have defects.In order to bridge these gaps,we proposed two adaptive clustering methods,which are density-based and gravity model-based insect pollination adaptive clustering algorithm.The density-based insect pollination adaptive clustering algorithm effectively converts the mapping mechanism of micro-macro behavior and nonlinear dynamics of adaptive Agent system to a mathematical model with theoretical basis,which exists in the the process of interaction between pollinators and plants.By using the mechanism of Agent pollinator's intelligent evaluation in the process of searching target object,the algorithm overcome the defect that large amount of computing resource has been wasted due to the low efficiency of the random search method.What's more,the sensitivity analysis of parameters is adopted in algorithm to avoid the defects caused by artificial experience setting.In addition,data objects oriented local and global strategies respectively improve the clustering compactness and the quality of the algorithm classification model.The similarity measure method based on Gaussian kernel function in algorithm will be updated in real time according to the results of the operation,which could find the basis for correctly dividing the data objects and adapt to the spatial structure characteristics of different datasets.The gravity model updates the position of target object to the high density neighborhood range according to the space distance.Based on the above fact,the evaluation mechanism of feature similarity between data objects is introduced into the gravity model.The improved gravity model uses feature similarity and position coordinates as weights to estimate the similarity and coordinate distance between data objects in the neighborhood range so as to obtain the new position.Moreover,the strategies that data oriented local and global updating make the new position of data object with higher fitness,which overcomes the defect that the traditional adaptive clustering algorithm just adopts the randomly moving rules and could not improve the fitness of data position.The algorithm is suitable for different types of datasets and is insensitive to noise.The effectiveness and performance of two proposed adaptive clustering algorithms based on insect pollination are verified by simulation experiments.The simulation results show that in the quality and stability of clustering results,the two adaptive clustering algorithms presented in this paper are generally superior to the current adaptive algorithms based on swarm intelligence.The feasibility and validity of the two algorithms achieve the goal of effectively converting the self-organizing and cooperative control strategies of biointelligent system into the framework of available artificial intelligence algorithms.