Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/513256Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Zulaiha Ali Othman, Assoc. Prof. Dr. | |
| dc.contributor.author | Ghassan Saleh Hussein Al-Dharhani (P62445) | |
| dc.date.accessioned | 2023-10-16T04:35:03Z | - |
| dc.date.available | 2023-10-16T04:35:03Z | - |
| dc.date.issued | 2017-06-16 | |
| dc.identifier.other | ukmvital:86194 | |
| dc.identifier.uri | https://ptsldigital.ukm.my/jspui/handle/123456789/513256 | - |
| dc.description | Climate pattern discovery from univariate data is a challenging process due to the similar data values that may exist throughout the year, as exemplified by the tropical climate data. Traditionally, climate time series are analysed statistically, but through this analysis, meaningful information is rarely obtained. In climate time series analysis, there are three major factors which influence the discovery of climate change pattern namely, (1) data representation, (2) data mining algorithms to extract patterns in the form of data mining algorithms, and (3) the methods used to visualize the patterns, which are meaningful and easy to interpret by domain experts. Improving these factors has a significant impact on the successful application of pattern analysis to climate change domain. Although several methods have been proposed in the literature, there are limitations as to their extraction of accurate and meaningful climate change patterns from univariate data. This research proposed three models of climate change patterns from such data. First, an enhanced Radial Basis Function Neural Network algorithm (ERBFN) was proposed based on back propagation to discover climate change Prediction Patterns (PP). Second, a fuzzy based symbolic data representation was proposed to discover climate change Shapelet Patterns (SP), known as Shapelet Patterns Algorithm (SPA), and third, climate change Frequent Patterns (FP) known as fuzzy based Temporal Frequent Pattern Algorithm (TFP) was proposed as enhanced Apriori algorithm. The proposed SPA discovered seven climate change patterns, which change over time in the form of coloured shapes to indicate annual changes in temperature patterns, such as cool, warm, hot and very hot, while the TFP is able to show how frequent climate pattern changes over time, and identify the appeared or disappeared patterns. TF pattern with specific time then obtained the similar climate change patterns as association rules mining among four stations. The experiments were conducted using four stations raw climate time series dataset that were collected from Petaling Jaya, Subang, KLIA Sepang and Universiti Malaya in Selangor. The result showed that the proposed E-RBFN neural network significantly obtained better prediction accuracy horizontally (average MSE 2.800850E-04) and vertically (average MSE 2.280075E-04) of all four stations compared with other neural networks such as Elman, NAR and original RBFN in both horizontal and vertical angles. SPA algorithm was able to visualize meaningful climate change patterns in the form of coloured shapes to indicate changes in temperature patterns. Finally, the TFP results have shown more meaningful patterns with performance in terms minConfidence measurement between 91% and 100% with an average of 98.4% compared with the original Apriori, Apriori-Total, and FP tree algorithms. The proposed work overcame various methods (E-RBFN, SPA, and TFP) to discover climate change pattern based on univariate data that helps climate change decision makers. Most climate experts who evaluated the patterns and accepted the results stated that all patterns produced meaningful information and presented detailed climate change informatics that can aid climate change experts in better decision making.,Ph.D. | |
| dc.language.iso | eng | |
| dc.publisher | UKM, Bangi | |
| dc.relation | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
| dc.rights | UKM | |
| dc.subject | Climate pattern | |
| dc.subject | Data representation | |
| dc.subject | Data mining algorithms | |
| dc.subject | Fuzzy based symbolic | |
| dc.subject | Universiti Kebangsaan Malaysia -- Dissertations | |
| dc.title | Time series prediction and fuzzy based algorithms for climate patterns discovery | |
| dc.type | Theses | |
| dc.format.pages | 192 | |
| dc.identifier.barcode | 002691(2017) | |
| Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ukmvital_86194+SOURCE1+SOURCE1.0.PDF Restricted Access | 276.13 kB | Adobe PDF |  View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.