Mongkon Youngtanurat.. Web service for latex trading data exchange and latex volume prediction. Master's Degree(Computer Engineering). Mahidol University. Mahidol University Library and Knowledge Center. : Mahidol University, 2016.
Web service for latex trading data exchange and latex volume prediction
Abstract:
This research aimed to develop a web service system that helps latex traders develop their own latex trading programs easily. They can exchange latex trading data via the provided web services. This data exchange could be beneficial to their business plans. Another aim of this research was to predict daily traded latex volume by using meteorological conditions and publish prediction results via a web service. Three classification methods were tested in order to select the most suitable one for building the prediction model, these were; decision tree, neural network, and support vector machine. The data used in this research was unbalanced. So the class distribution was adjusted by sampling with replacement and bias toward uniform distribution. Results from experiments showed that decision tree gave higher overall accuracy than the others. From class-by-class analysis, it was also robust to the overlearning of small classes. By increasing the volume of the smallest class, it improved the prediction of this class, yet kept the prediction of the others stably well.