000			04734ntm a2200373 i 4500
999			_c91145 _d91151
003			MY-KuUP
005			20251125105425.0
006			t||||fr|||| 000 0
007			ta
008			191112t20192019my a|||frm||| 000 0 eng d
020			_aTHE0008331(Local) _qhardback
040			_aUMP _beng _cUMP _erda
090			_aFSKKP .D43 2019 r Thesis
100	1		_aDas, Debashish, _eauthor.
245	3	0	_aAn ensemble of neural network and modified grey wolf optimizer for stock prediction / _cDebashish Das
264		1	_aKuantan, Pahang : _bUMP, _c2019
264		4	_c© 2019
300			_axiii, 155 pages : _billustrations ; _c30 cm. + _e1 CD-ROM
336			_atext _2rdacontent
336			_atext _2rdacontent
337			_aunmediated _2rdamedia
337			_acomputer _2rdamedia
338			_avolume _2rdacarrier
338			_acomputer disc _2rdacarrier
347			_atext file _bPDF _2rda
500			_aFaculty of Computer Systems and Software Engineering
502			_aThesis (Doctor of Philosophy) -- Universiti Malaysia Pahang – 2019
504			_aIncludes bibliographical references
520	3		_aOptimization relates to the process of finding the optimum solution (either maximize or minimize) to a particular problem satisfying some given constraints. Owing to its simplicity and flexibility, meta-heuristics have been proven to be effective for solving optimization problems. To date, there are many meta-heuristics have been developed in the literature. In line with the No Free Lunch theorem which suggests that no single metaheuristic is the best for all optimization problems, the search for better algorithms is still a worthy endeavour. Grey Wolf Optimizer (GWO) is a recently developed meta-heuristic algorithm which is appealing to researcher owing to its demonstrated performance as cited in the scientific literature. Despite its performances, GWO is not without limitation.Precisely, the current best optimal individual of GWO is biased toward alpha and other individuals (e.g. beta and delta) attempt to modify their positions toward this best individual in each iteration process. This update process may cause the algorithm to fall to local optima especially in the cases where there are many competing local optima. Therefore, the research attempts to modify GWO to addresses the limitation of GWO for improvement of exploration by strengthen the searching process via several random leaders in each iteration, re-generating the random leaders in each iteration and introducing archive to verify the solution with better probability to proceed further for training and re-generation. The verification of each solution individually by Modified GWO, instead of considering as a final solution, facilitates the improvement of the exploration. Additionally, the research restricts the number of variables through feature selection to enhance the performance of the algorithm. Subsequently, the research attempts to construct an ensemble model applying Modified Grey Wolf Optimizer (MGWO) and neural network for stock prediction. Widespread models like Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Ant Colony Optimization (ACO), Evolutionary Strategy (ES) and Population-Based Incremental Learning (PBIL) dealing with the specified problems are also explored and compared. The research implements stock prediction analysis as a case study for training the neural network by adopting MGWO algorithm. In this research, data is collected from reputed stock markets; New York Stock Exchange (NYSE), NASDAQ and emerging markets; Dhaka Stock Exchange (DSE), Bursa Malaysia. Moreover, various factors data like Dollar price, Gold price, Bank interest rate, Foreign Direct Investment, and Inflation are collected to measure the effect in stock market. K-means clustering is applied to select the highly promising company; MGWO is implemented for feature selection and training; finally, MGWO-NN is applied to predict the stock price. The “ensemble” model selected here to achieve better predictive performance, is used to predict future market price. The proposed approachoutperforms existing available meta-heuristic algorithms. Specifically, the proposed model achieved 97% classification rate, 95% precise prediction and less than 2.0 error rate. In conclusion, the successful implementation of MGWO and ensemble model makes a valuable contribution to scientific arena
610	2	0	_aFaculty of Computer Systems and Software Engineering _xDisertations
650		0	_aUniversities and colleges _xDisertations
650		0	_aTheses
942			_2lcc _cTHESIS