Developing An Expert System To Forecast The Stock Market by Mike Flanagan, Ph.D.
Nowadays, computers are often used to develop rule-based trading strategies. Frequently, these rules will be based on work provided by experts; such rule-based models are generally called expert systems. Here, Mike Flanagan takes you through the process of developing an expert system for generating medium- (three- to six-month) and long-term (six- to 12-month) forecasts of the US stock market.
An expert system is a computer program that uses deductive logic to simulate the decision-making process of humans. The program contains a modifiable knowledge base consisting of if-then-else rules and facts, a user interface, a database/spreadsheet interface and an inference engine that makes logic-based decisions. In the early days of expert system development, systems were built by interviewing a human expert about a particular subject and then attempting to capture that expert's knowledge for the benefit of other, nonexpert users; the term expert system is derived from this process. Today, the goal of an expert system developer is to capture as much relevant knowledge about a given subject from all
available sources and create a series of rules that represent this knowledge. The performance of an expert system model is primarily a function of the quality and size of its knowledge base.
The if part of a rule is basically a series of conditions expressed as grammatical sentences or algebraic expressions. If the program determines that the if conditions are true, the statements in the then part are assumed to be true and are added to the knowledge base as known facts. If, however, any of the conditions in the if part are not true, the statements contained in the else part are assumed to be true and this information is also added to the knowledge base.