Please use this identifier to cite or link to this item: https://ptsldigital.ukm.my/jspui/handle/123456789/476349
Title: Optimizing adaptive neuro-fuzzy inference system using bees algorithm for stock market prediction
Authors: Mohammad Mahdiani (P65695)
Supervisor: Mohd Zakree Ahmad Nazri, Assoc. Prof Dr.
Keywords: Stock market
Neuro-fuzzy inference system
Bees algorithm
Dissertations, Academic -- Malaysia
Issue Date: 4-Mar-2015
Description: Stock market prediction method is important and of great interest to the computer science community because an effective method may promise attractive benefits to the investors around the world. Prediction is defined as a process which project the future performance according to the historical information which exists. However, the stock market is always difficult to accurately predict due to many reasons, such as the political situation (for some specific countries), the global economy, etc. Stock price prediction is as one of the most challenging applications of time series forecasting. Unfortunately, there are some drawbacks to the conventional time series models: the rules mined from genetic algorithms and artificial neural networks are not easily understandable, and most rule-based forecasting models generate too many predicting rules for a stock price index and stock market investors usually make short-term decisions based on recent price fluctuations, but most time series models use only the last period of stock market price in forecasting, and also there is no research that has used ANFIS model to predict the proposed stock markets which we used in this research. The advantage of neural network that have been applied in many researches are capablity of handling noisy measurements and require no assumption about the statistical distribution of the monitored data. This thesis has 3 main objectives. The first objective is to investigate the adequacy of neural networks model and ANFIS for stock market prediction. After we have investigated the effectiveness of ANFIS in stock price prediction, our second objective is to improve ANFIS. The suggested method includes two main modules: the prediction module and the optimization module. In the prediction module, adaptive neuro-fuzzy inference system (ANFIS) is investigated. In ANFIS training, the vector of radius has very important role for its recognition accuracy. Therefore, in the optimization module, Bees algorithm (BA) is proposed for finding the optimum vector of radius. Bees algorithm is one of the most recent and powerful algorithm based on collective intelligence and its efficiency in optimization has been proven. Our third objective is to investigate which membership function in the fuzzy inference engine produces the best result. The proposed method uses fuzzy rules for recognition task. The crucial input variables are chosen by this method while it is building the fuzzy model by combining a least squares approximation algorithm with cluster approximation method from available data. These variables include the previous day’s cash market high, low and closing prices and today’s opening cash index. These investigations produce new knowledge on the effectiveness of ANFIS based algorithm for stock price prediction. Furthermore, the effect on using different membership functions in stock price prediction gives valuable trade-off information about each membership function. Simulation results show that the proposed system has higher prediction accuracy in comparison to the conventional mode. The experiment results of this studies shows that the hybrid ANFIS with Bees algorithm performs better than the ANFIS. The results indicate that an optimized ANFIS can be a useful tool for stock price prediction in emerging markets, especially the Asian stock market.,Certification of Master's/Doctoral Thesis" is not available
Pages: 131
Publisher: UKM, Bangi
URI: https://ptsldigital.ukm.my/jspui/handle/123456789/476349
Appears in Collections:Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Files in This Item:
File Description SizeFormat 
ukmvital_82211+SOURCE1+SOURCE1.0.PDF
  Restricted Access
269.94 kBAdobe PDFThumbnail
View/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.