A Comparison Of Type-1 Fuzzy And Type-2 Fuzzy Methods In Anfis Modelling To Analyze FTSE Bursa Malaysia Kuala Lumpur Composite Index

Lee, Kin Fei (2013) A Comparison Of Type-1 Fuzzy And Type-2 Fuzzy Methods In Anfis Modelling To Analyze FTSE Bursa Malaysia Kuala Lumpur Composite Index. Project Report. UTeM, Melaka, Malaysia. (Submitted)

	PDF (24 Pages) A_Comparison_Of_Type-1_Fuzzy_And_Type-2_Fuzzy_Methods_In_Anfis_Modelling_To_Analyze_FTSE_Bursa_Malaysia_Kuala_Lumpur_Composite_Index.pdf - Submitted Version Download (574kB)
	PDF (Full Text) A_Comparison_Of_Type-1_Fuzzy_And_Type-2_Fuzzy_Methods_In_Anfis_Modelling_To_Analyze_FTSE_Bursa_Malaysia_Kuala_Lumpur_Composite_Index.pdf - Submitted Version Restricted to Registered users only Download (1MB)

Abstract

Many investors have been using technical analysis in stock market prediction. However, technical indicators are facing difficulties in handling uncertainty and stochastic factors. This has encouraged the research of soft computing techniques in stock data analysis on the features extracted using the technical indicators. ANFIS and its variations are widely applied model in stock market prediction due to its ability in determining converge time during data learning process. Research work on Type-2 Fuzzy inference had proven to perform better than conventional Type-1 Fuzzy inference in many common dataset due to its interval membership functions can well model uncertainty than Type-1 Fuzzy. However, there is no convincing comparison on real-life stock data especially in FTSE Bursa Malaysia Kuala Lumpur Composite Index (KLCI). Hence, a comparison of Type-1 Fuzzy and Type-2 Fuzzy in ANFIS modelling is necessary typically on the FTSE Bursa Malaysia KLCI data to provide a more in depth analysis on the different fuzzy inference method in ANFIS modelling. The project experimented on the Type-2 Fuzzy interval ranges from 0.01 to 0.10 with step size of 0.01 to demonstrate the effect of Footprint of Uncertainty (FOU) range towards prediction performance. Experiment is conducted in Matlab R2011b environment. The results are compared in terms of Root Mean Square Error and validated using two tailed t-test. Experiment results shows that overall performance of Type-2 Fuzzy are better than Type-1 Fuzzy in ANFIS modelling due to its interval membership functions to handle stock uncertainty, where the accuracy showed an increasing trend from FOU range 0.01 to 0.10. However, this experiment is not able to suggest the best FOU range. Hence, optimizing FOU range is suggested for future work to achieve better prediction accuracy.

Actions (login required)

View Item