(IJCSIS) International Journal of Computer Science and Information Security, Vol. 14 No.3, March 2016 Performance Comparison between Forward and Backward Chaining Rule Based Expert System Approaches Over Global Stock Exchanges Sachin Kamley Deptt. of Computer Application’s S.A.T.I., Vidisha, India Shailesh Jaloree Deptt. of Appl. Math’s and CS S.A.T.I., Vidisha, India R.S. Thakur Deptt. of Computer Application’s M.A.N.I.T., Bhopal, India Abstract— For the last couple of decade’s stock market has been considered as a most noticeable research area everywhere throughout the world because of the quickly developing of the economy. Throughout the years, a large portion of the researchers and business analysts have been contributed around there. Extraordinarily, Artificial Intelligence (AI) is the principle overwhelming area of this field. In AI, an expert system is one of the understood and prevalent techniques that copy the human abilities in order to take care of particular issues. In this research study, forward and backward chaining two primary expert system inference methodologies is proposed to stock market issue and Common LISP 3.0 based editors are used for designing an expert system shell. Furthermore, expert systems are tested on four noteworthy global stock exchanges, for example, India, China, Japan and United States (US). In addition, different financial components, for example, Gross Domestic Product (GDP), Unemployment Rate, Inflation Rate and Interest Rate are also considered to build the expert knowledge base system. Finally, experimental results demonstrate that the backward chaining approach has preferable execution performance over forward chaining approach. Keywords— Stock Market; Artificial Intelligence; Expert System; Macroeconomic Factors; Forward Chaining; Backward Chaining; Common LISP 3.0. I. INTRODUCTION In nowadays stock price prediction has become sizzling topic in the time series analysis and dependably stays in the limelight because of rising and falling states of the economy. So different researchers and business experts have paid consideration to break down and anticipating the future estimation of stock exchange prices [1] [2]. Previously, the various tools and techniques have been proposed for the numeric stock value prediction. However, an Artificial Neural Network (ANN) is one of the most prominent techniques among them. For most recent five decades an expert system has risen as effective AI techniques and had demonstrated to its value in different areas such as designing, account, farming, medicine, crystal gazing and numerous more [3]. An expert system is likewise called knowledge based system which uses knowledge to tackle problems and knowledge must be encoded in some forms of facts, rules, procedures, relations, etc. In expert system knowledge might be gathered from different sources, for example, primary source human expert and secondary sources, for example, books, magazines, newspapers, reputed journals [4]. After knowledge acquisition, its representation is essential and very challenging issues. For symbolic representation of knowledge, there are various techniques are utilized such as frames, semantic net, scripts, production rule and so on. Figure 1 shows the basic structure of expert systems. Knowledge Base Database User User Interface Explanation Mechanism Inference Engine Rule : IF- THEN Facts Knowledge Engineer Expert Knowledge Fig. 1. Basic Structure of an Expert System [3] [5] In Figure 1 inference engine is the most capable component of the expert system that is in charge of completing the reasoning procedure where the expert system comes to end at specific solution. Additionally, the inference engine is likewise responsible for links the rules with the knowledge base where facts are given by the database. Throughout the previous couple of years, stock market plays an important role in the quickest developing economy, but accurately shares price forecasting are still very chaotic and complicated process [6]. There are various algorithms are developed for numerical forecast, but no authors have endeavored symbolic representation of stock market data. There are two basic approaches forward chaining and backward chaining are utilized as a part of the expert system design and development. Forward chaining is a data driven methodology which begins from the known data and continues forward with that https://dx.doi.org/10.6084/m9.figshare.3153883 74 https://sites.google.com/site/ijcsis/ ISSN 1947-5500