Artificial Bee Colony Optimization Perception

We have entered the era of big data,which is a time of big bang for available information and provides an excellent opportunity for the development of modern society.In the future,the core competitiveness of industries largely depends on the speed and ability to convert big data into information and knowledge,which depends on the level of data analysis and application.In order to gain competitive advantage for any country or region,big data analysis and mining is a highly strategic field.However,due to the high dimensional characteristics of big data and large sample,data analysis is facing great challenges,which is shown in three respects: First,high dimensional characteristics bring noise pollution,pseudo correlation and non-homogeneity;Second,high dimensionality and massive sample lead to algorithm instability and huge computational cost;Lastly,when using different processing methods to deal with the massive sample of multisource aggregation at different time points,it brings some hotspot issues such as non-homogeneity,experimental variation and statistical bias.In order to deal with the challenge of big data,the existing statistical methods and calculation methods are no longer applicable,new statistical methods and stable fast algorithms are needed.Many machine learning algorithms are used for big data analysis and processing,such as support vector machine for customer classification in bank,telecom and enterprise,which have avoided the difficulties caused by the high dimensional characteristics of data.The idea of support vector machine comes from the perceptron,which learns a separation hyper-plane from the training data set that is divided into positive and negative classes.However,because of that the perceptron iterative algorithm is related to the selection of the initial iteration point and the iteration termination condition,the learning separation hyper-plane is not unique,which lead to that the generalization ability of perceptron is not too good.In this paper,the swarm intelligence algorithm is introduced into the perceptron learning algorithm,constructing the following iterative loss function of the perceptron:The loss function reflects the number of misclassified points.Consistent with the beecolony algorithm,in this paper,the minus iterative loss function is taken as the fitness function of bee colony algorithm.The process of optimizing the separation hyper-plane by the swarm algorithm is to maximize the fitness function,alasThe experimental results show that the classification results obtained by the proposed algorithm are better than the traditional perceptron.