期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2016
卷号：92
期号：2
出版社：Journal of Theoretical and Applied
摘要：Rainfall has influence the social and economic activities in particular area such as agriculture, industry and domestic needs. Therefore, having an accurate rainfall forecasting becomes demanding. Various statistical and data mining techniques are used to obtain the accurate prediction of rainfall. Time series data mining is a well-known used for forecast time series data. Therefore, the objective of this study is to develop a distribution of rainfall pattern forecasting model based on symbolic data representation using Piecewise Aggregate Approximation (PAA) and Symbolic Aggregate approXimation (SAX). The rainfall dataset were collected from three rain gauge station in Langat area within 31 years. The development of the model consists of three phases: data collection, data pre-processing, and model development. During data pre-processing phase, the data were transform into an appropriate representation using dimensional reduction technique known as Piecewise Aggregate Approximation (PAA). Then the transformed data were discretized using Symbolic Aggregate approXimation (SAX). Furthermore, clustering technique was used to determine the label of class pattern during preparing unsupervised training data. Three type of pattern are identified which is dry, normal and wet using three clustering techniques: Agglomotive Hierarchical Clustering, K-Means Partitional Clustering and Self-Organising Map. As a result, the best model has be able to forecast better for the next 3 and 5 years using rule induction classification techniques.
关键词：Time Series Data Mining; Clustering; Classification; Time Series Symbolic Representation; Rainfall Forecasting
