Sureeluk Ma. Predicting Drought Indices in Nakhon Ratchasima Province using a Deep Belief Network with Restricted Boltzmann Machines. Master's Degree(Applied Mathematics). Prince of Songkla University, Pattani Campus. Office of Academic Resources. : Prince of Songkla University, Pattani Campus, 2018.
Predicting Drought Indices in Nakhon Ratchasima Province using a Deep Belief Network with Restricted Boltzmann Machines
การทำนายดัชนีความแห้งแล้งในจังหวัดนครราชสีมาโดยใช้ Deep Belief Network กับ Restricted Boltzmann Machines
Abstract:
In this study, we examine the ability of deep learning in making prediction from time series data. First, the precipitation data from Nakhon Ratchasima province in northeastern region of Thailand is converted into various types of standardized precipitation index (SPI). Next, for each SPI, a deep belief network, consisting of restricted Boltzmann machines, learns its parameters from data through unsupervised path using minimized contrastive divergent algorithm follow by supervised path using backpropagation algorithm. Last, the prediction accuracies from all types of the standardized precipitation index are evaluated and compared. The result shows that the long term SPI of 12 months makes more accurate prediction than the short term SPI of 3, 6, and 9 months.