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

Title:	Model peramalan taburan hujan menggunakan teknik perlombongan data siri masa
Authors:	Ommi Rahmah Abdullah (P42893)
Supervisor:	Zulaiha Ali Othman, Prof Madya Dr
Keywords:	taburan hujan Data mining
Issue Date:	15-Aug-2011
Description:	Air merupakan sumber semulajadi yang penting terutama dalam sektor pertanian, sektor industri dan keperluan harian manusia. Meramal taburan hujan yang tepat amatlah sukar. Pelbagai teknik statistik dan perlombongan data telah dipraktikkan untuk mendapatkan nilai peramalan yang tepat di masa hadapan. Sehubungan itu, objektif kajian ini adalah untuk membangunkan model peramalan taburan hujan mengguna teknik perlombongan data siri masa ke atas data siri taburan hujan di kawasan Langat, Selangor. Pembangunan model terdiri daripada empat fasa utama iaitu pengumpulan data, prapemprosesan data, pemodelan dan penilaian. Model dibangunkan berasas data taburan hujan harian pada bulan Mei-September (147 hari) untuk jangka masa 41 tahun (1953-1993) dari stesen 17R-JPS Ampang. Prapemprosesan data adalah proses yang paling penting dalam perlombongan data siri masa. Proses utama diperingkat ini menukarkan nilai data asal kepada nilai perwakilan siri masa menggunakan teknik pengurangan dimensi PAA (Piecewise Aggregate Approximation) dan pendiskritan SAX (Symbolic Aggregate Approximation). Proses ini mengurangkan saiz dimensi corak asal (147 hari) kepada saiz dimensi yang lebih pendek (21 minggu). Manakala teknik pengelompokan Self-Organizing Map (SOM) digunakan untuk mengenal pasti corak taburan hujan kering dan normal. Pembangunan model peramalan mengguna teknik pengelasan pohon keputusan J48, pengelasan petua JRip, dan teknik regresi linear. Keputusan uji kaji menunjukkan model peramalan teknik pengelasan menghasilkan ketepatan ramalan yang konsisten berasaskan keputusan purata peratusan ketepatan (81.55% - 95.76%), purata bilangan petua (2-5), dan purata nilai RMSE (0.18 - 0.36). Keputusan uji kaji juga menunjukkan nilai RMSE model pengelasan (0.0-0.66) lebih kecil berbanding model regresi linear (3.46-6.37) untuk jangka masa setahun hingga sepuluh tahun ke depan, bila diuji dengan set data baru yang diperolehi dari 5 stesen di sekitar kawasan Langat. Keputusan ini menunjukkan bahawa model peramalan pengelasan menggunakan teknik perlombongan data siri mampu meramal dengan tepat taburan hujan di kawasan yang berhampiran dengannya adalah boleh diterima.,Water is the most important a natural resources, especially in agriculture sectors, industry sectors, and it is essential to human daily life. To forecast the accuracy of rainfall is very difficult. Various statistical and data mining techniques are used to obtain the accurate prediction in the future. Consequently, the objective of this study is to develop a rainfall forecasting model using time series data mining techniques against rainfall series data in Langat area, Selangor. The development of the model consists of four main phases: data collection, data pre-processing, modelling and evaluation. The model is built based on daily rainfall data from May-September (147 days) for a period of 41 years (1953-1993) from the 17R-JPS Ampang station. Data pre-processing is the most important processes in the time series data mining. The main process at this stage to transform the raw data to the time series represent using dimensional reduction PAA (Piecewise Aggregate Approximation) and SAX (Symbolic Aggregate Approximation) discretization. This process is reducing the raw patterns dimension size (147 days) to the shortest dimension size (21 weeks). Meanwhile, Self-Organizing Map (SOM) clustering technique is used to determine dry and normal rainfall patterns. A forecasting model was built using the decision tree classification J48, the classification rule JRip, and the linear regression techniques. The results shows forecasting model classification techniques produces the accurate prediction consistently based on average percentage of accuracy result (81.55%-95.76%), average number of the rules (2-5), and average value of RMSE (0.18-0.36). The result shows the RMSE value of classification model (0.0–0.66) is smaller than the linear regression model (3.46-6.37) for a period of one to ten years later, when it was tested with the new dataset which obtained from 5 stations around Langat area. These results indicate that the classification model using time series data mining technique able to forecast accurately the rainfall in the nearby area can be accepted.,Master/Sarjana
Pages:	117
Call Number:	QA76.9.D343O535 2011 3 tesis
Publisher:	UKM, Bangi
URI:	https://ptsldigital.ukm.my/jspui/handle/123456789/476371
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.