The Prediction Of Sailing Speed Based On Support Vector Machine

Sailing is a kind of sports which relies on natural wind force acting on sail, also, Sailing counts on athletes manipulating sail and adjusting the gestures of sail. So, it’s a sport which involves in competition, sight enjoyment, venture, entertainment etc. The result of Sailing competition is subject to multiple factors, like sea environment (sea wind, ocean current), athletes’ manipulation skills and sailing course selection strategies, in some degree, it’s a synthetic contest among stamina, skills and wisdom. Undoubtedly, it would be a valuable rese’arch with theoretical significance and application value if we can apply scientific methods into analyzing and predicting Sailing results to supervise daily training and improve competition results.Owing to the complexity, time variant and indetermination of Sailing, it’s hard to depict the motion raw accurately and genuinely with traditional kinematic dynamics analysis methods. Based on series of collected data from earlier research project named Sailing Environmental Monitoring and Motion Analysis system development, a Sailing database is set. So, in this paper, we propose to set up a prediction model of Sailing velocity time series via scientific methods, like, Data Mining or Machine Learning. Moreover, this model can be viewed as the theoretical basis to predict and assess VMG (Velocity Made Good, VMG) which is the effective velocity in the direction of destination when sail upwind. Expectedly, we can provide scientific analysis method for Sailing technology diagnosis.In fact, the problem of Sailing velocity prediction can be treated as a typical time series prediction problem influenced by natural environment, sailboat and individuals, with the characteristic of nonlinear and small sample. As a new Machine Learning method, SVM (Support Vector Machine, SVM) has its unique advantage over solving the small sample and nonlinear problems, and it has been widely used to solve pattern recognition and nonlinear regression problems with high precision, fast convergence and simple operation characteristics. Based on the actual collected sample data on Sailing, in this paper, we apply SVM in the prediction of sailing velocity, in other words, this paper is aimed at achieving the prediction of Sailing motion parameters. Apart from the prediction of sailing velocity (magnitude, direction), we also compute and analysis the VMG which can assess the effective velocity in the direction of destination.In this paper, firstly, we introduce the preprocessing of the collected data of Sailing and the process of setting up corresponding data warehouse. Above all, this paper proposed a new Sailing velocity prediction method based on Support Vector Machine, also, make comparisons with BP (Back propagation, BP) neutral network in terms of learning performance. Afterwards, we explore a method based on fuzzy set to granulate the collected data, then use SVM to make prediction for the fuzzy information granule, finally, we achieve the effective prediction about future trends and changes of the sailing velocity and VMG. At last, we use real Sailing data as simulation samples to do data experiment. Simulation results show that SVM based on fuzzy information granulation can achieve accurate prediction about the future trends of sailing velocity and VMG.