The Research And Implementation Of The Prototype System Of The Support Vector Machine Based On Storm In The Forecasting Of Logistics Shipping Index

In recent years, with the rapid development of information technology, modern logistics industry has entered the era of information technology. International shipping index BDI(Baltic Dry Index) is the economic indicator of the shipping industry, which contains the shipping industry’s dry bulk trading volume change and combines the characteristics of the future economic activities with the leading economic indicators. However, the influence factors of BDI are very complex, and it is a nonlinear complex system. The traditional forecasting method is based on the accurate mathematical model, but the effect is not ideal. Therefore, it is extremely important to investigate the internal law and external influence of freight index fluctuation, and develop a new model based on artificial intelligence to forecast the freight index, which is very important for us to grasp the market form and avoid the risk. So this paper summarizes the BDI model of forecast research and analysis the factors affecting the change of BDI and systematically study the online learning, parallelization, and application in BDI prediction.Firstly, an improved on-line support vector machine regression learning algorithm(IOSVR) is proposed. In the face of challenge of growing information, IOSVR based on support vector machine uses dynamic adaptive adjustment method, able to work in a dynamic process of adaptively adjusting the training parameters, optimization iterative training procedure to obtain better prediction precision and generalization ability. The simulation results show that IOSVR can not only ensure the accuracy of the training, but also significantly reduce the learning time and the classification error rate. So IOSVR is more suitable for the modeling of the actual process.Secondly, a BDI prediction method based on support vector machine combining with the characteristics of BDI is proposed. This paper deeply analyzes the supply and demand market of international shipping, combining with the characteristics of the Chinese market, discusses the influencing factors of BDI, and selects the main characteristic variables and latent variables of BDI. Then the BDI prediction model based on IOSVR is proposed and studied. This model fully considers the randomness of the freight index. Through experiments, it shows that the model can better predict the BDI, can better describe the change trend of BDI, and is more suitable for the prediction of nonlinear time series.Finally, the parallelization of IOSVR in the storm platform combined with the storm parallel mechanism is achieved. With increasing of the size of datasets, in order to improve the data processing ability and the running efficiency of the serial support vector machine algorithm. The parallel learning of IOSVR in the storm platform based on the characteristics of storm is realized, which integrate the advantages of each node in parallel computing effectively and also provides the reliable data storage and the reliable processing capability. In addition, the IOSVR based on storm is applied to the BDI prediction system, which provides a powerful tool for logistics management.