Research On Logistics Data Classification Algorithm Based On Incremental Learning

Logistics industry is developing rapidly with the progress of science and technology.In order to preempt the logistics market,logistics enterprises make full use of modern advanced technology,such as data mining,cloud computing,Internet of Things.These modern technologies not only bring great economic benefits for the logistics enterprises,but also accelerate the development of the domestic logistics.Since the rapid development of the logistics industry produces a large number of logistics-related data,how to deal with these data efficiently has become a hot topic to be addressed for today’s logistics industry.In this thesis,we mainly focus on the process of massive logistic data.Specifically,the classification algorithm Support Vector Machine(SVM)is improved,which is suitable for logistic data processing.This algorithm can improve the processing speed of logistics industry-related data and reduce the data redundancy.Besides,the algorithm can mine potential value from the data and provide the decision support for logistics enterprises,which can improve the low efficiency issue of the traditional logistics data processing.The main contributions of this thesis can be summarized as follows:(1)In order to slove the problem of slow convergence speed in logistics data classification processing,this thesis first studies the stochastic gradient descent algorithm(SGD).For the problem that gradient descent direction is not the optimal solution direction when SGD is used to search global optimization solution,a new Stochastic Gradient Descent algorithm based on Double Sample(DSSGD)is proposed.DSSGD algorithm takes the resultant vector of the current sample gradient and the previous sample gradient as the gradient descending direction,which can optimize the data convergence direction and improve the logistics data processing speed.Simulation results show that proposed DSSGD algorithm can effectively optimize the gradient descent convergence direction,and improve the processing speed and accuracy.(2)In view of the serious historical data redundancy problem in logistics data processing,the thesis proposes an incremental learning classification algorithm for forgetting factors based on stochastic gradient descent support vector machine algorithm,i.e.,Stochastic Gradient Descent Support Vector Machine Based on Forgetting Factor α(α-SVMSGD).This algorithm adds the forgetting mechanism to the classification algorithm,and selectively removes and preserves the historical data according to the forgetting factor value of the data samples after several training,so as to reduce the logistics data redundancy.In addition,an adaptive adjustment forgetting factor mechanism is proposed,which can make the classifier constantly accommodate the incoming logistics data,find new data classification,and mine useful information.Simulation results show that the α-SVMSGD algorithm can effectively reduce the data redundancy and improve the classification speed.According to the forgetting factor value,α-SVMSGD classifier can fit to the real-time updated logistics data and discover new data classification.