1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Neural network variants for Malaysian weather prediction
Authors:	Siti Nur Kamaliah Kamarudin (P61047)
Supervisor:	Azuraliza Abu Bakar, Professor Dr.
Keywords:	Neural networks (Computer science)
Issue Date:	5-Jul-2013
Description:	Weather prediction has been an active research since many years, and recently the issue of global climate change has gained the interest of researchers to further explore this problem. Many data mining algorithms based on neural network have been developed to solve the weather prediction problem specifically the rainfall prediction. The main issues addressed in rainfall prediction are the large amount of weather data stream and the accuracy of prediction. In order to develop a good prediction model, two factors need to be considered; (i) representation of the time series data, and (ii) achieving high predictive accuracy. This study aims to investigate the performance of Back Propagation Neural Network (BPNN) algorithm variants for the Malaysian weather prediction problem. In this study, five Neural Network variants are experimented along with hybrid neural network algorithms which include the standard BPNN, BPNN with Momentum and Quick-Propagation, Genetic Algorithm with Neural Network (GA-NN) and Particle Swarm Optimization with NN (PSO-NN). The experiments follow the standard data mining steps which consist of data collection and preparation, development of model, testing and evaluation. The original Malaysian rainfall dataset are obtained from the Institute of Climate Change, UKM which consists thirty years of data collected daily from ten stations. The Malaysian rainfall data are represented using a symbolic based representation technique conducted by previous researchers. The symbolic data are classified into four classes namely no rain, light, moderate, and heavy rain. The experimental results show that standard BPNN achieved the best prediction accuracy as compared to other algorithms.,Master / Sarjana
Pages:	89
Call Number:	QA76.87.S578 2013
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.