Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/513211Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Salwani Abdullah , Prof. Dr. | |
| dc.contributor.author | Jaddi Najmeh Sadat (P59630) | |
| dc.date.accessioned | 2023-10-16T04:34:43Z | - |
| dc.date.available | 2023-10-16T04:34:43Z | - |
| dc.date.issued | 2014-10-20 | |
| dc.identifier.other | ukmvital:82111 | |
| dc.identifier.uri | https://ptsldigital.ukm.my/jspui/handle/123456789/513211 | - |
| dc.description | In recent years, methodically expansion of database and information technology was continued using a more complicated and intelligent process to achieve the goal of any data mining tasks. Intelligent systems provide a great potential for advanced and theoretical research in data mining area. Classification and time series prediction are two important tasks in data mining. Based on literature review, optimisation of Artificial Neural Network (ANN) model using population-based meta-heuristic algorithms are good approaches with high level of achievements to handle classification and time series prediction problems. However, there are still a lot of rooms for improvement. Some of the important challenges include finding an appropriate solution representation; discover optimisation approach for a better tradeoff between exploration and exploitation in avoiding an early convergence and getting trapped in a local optimum, maintaining the diversity of the solutions in the population and investigating the applicability of the proposed approach for real world data. The research work presented in this thesis aims to build upon the state of the art in search methodologies for classification and time series prediction. Towards this aim, several objectives are introduced to increase the overall performance of classification and time series prediction approaches. The research first highlights an initial study on a genetic algorithm-based approach with a new solution representation to optimise both structure and weights of ANN model in obtaining better performance (in terms of accuracy) with less processing time. The promising results obtained from two categories of standard datasets i.e. six classification and two time series prediction are a motivation to next investigate a new population-based approach (bat algorithm, BA, in this case). Five variants of bat algorithms with optimised parameters setting based on Taguchi method incorporate with a personal best in velocity adjustment and chaotic maps are experimented in order to find a balance between the exploration and exploitation process, overcome the premature convergence and avoid from easily trapped into a local optimum. Statistical analysis shows that the variant of bat algorithm with personal best and Logistic map outperforms other variants (coded as TM2LogBat). Next, in maintaining the diversity of the solutions in the population, three existing cooperative topologies (Ring, Master-Slave and Coevolving) and two hybrid cooperative topologies (combination between Ring and Master-Slave, and Coevolving and Master-Slave) between sub-populations are investigated on the T-M2LogBat approach. The method with the combination of the Ring and Master-Slave (coded as T-M2Log-RMSBat) has better performance compared to other cooperative strategies. The performance of these methods is finally tested on rainfall real-world time series data taken from state of Selangor in Malaysia. Overall comparison indicates that TM2Log-RMSBat works well across standard datasets and able to obtainseveral best results in comparison with the state of the art, and also shows good performance on real-world datasets.,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 | Time series analysis. | |
| dc.title | Modified bat algorithm and artificial neural network for classification and time series prediction | |
| dc.type | Theses | |
| dc.format.pages | 204 | |
| dc.identifier.callno | QA280 .J334 2014 3 | |
| dc.identifier.barcode | 001750 | |
| Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ukmvital_82111+SOURCE1+SOURCE1.0.PDF Restricted Access | 3.49 MB | Adobe PDF |  View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.