Neuro-fuzzy techniques for financial engineering

Soft computing has been increasingly popular in many industrial and real-life applications. This project covers one aspect of soft computing; Neuro-Fuzzy techniques applied in financing engineering. Detailed research works and literature reviews are done in order to grasp Neuro-Fuzzy concepts and its applications. A real-life problem has been derived to find out whether news impact on the Singapore stocks in the SGX market. The Factiva database is used to search for news data, Yahoo! Finance for stock prices and Matlab software for programming. Two stocks namely, OCBC and DBS are compiled and used to input and train the created Neuro-Fuzzy program. Two types of encoding are used which are Binary Coding Method and Penta Coding Method (BCM & PCM). RMSE of 0.2514 and 3.0761 are achieved from the output of program. From this result, it reflects mixed responses. The value 0.2514 reflects a reasonable level of accuracy and 3.0761 reflects a lower level of accuracy. One possible reason deduced for the discrepancy could be due to misinterpretation of news encoding into numerical values. Limitations mentioned in the report posed problems to complete this project. Suggestions like having a two people project to handle different tasks will not only improve the efficiency of the tasks completion but will also improve the consistency and accuracy of the data encoding of news headlines into numerical values for training values. Overall, it is relatively successful having to complete the objectives of the project despite discrepancies of the RMSE values of the two stocks.Bachelor of Engineerin