Continue in 2 seconds

MEDai Generates Significant Return on Stock Portfolio with WizWhy

Published
  • July 01 1999, 1:00am EDT

BACKGROUND: MEDai, located in Orlando, Florida, uses a combination of artificial intelligence and statistical analysis to provide solutions for the healthcare industry. These solutions include processes for clinical decision support, high-risk prediction and provider profiling. MEDai also uses its artificial intelligence technology to predict the movement of stocks and has had tremendous success in this area.

PLATFORMS: Pentium II 400 MHz PC with 256MB RAM, Windows NT.

PROBLEM SOLVED: MEDai was looking for a way to predict the stock market using chart pattern recognition indicators. Neural net technology could not make much of these patterns because they occurred with a very low frequency, the patterns were very subtle and the returns of stocks had such a high standard deviation. Using WizWhy, we were able to find these patterns and generate a significant return on our stock portfolio.

PRODUCT FUNCTIONALITY: The latest version of WizWhy (Version 2.01) is a big improvement over previous versions. It is easy to walk through the predictive modeling process, which includes loading the data, selecting important parameters, generating the rules, saving the model and applying the model to a validation data set. It does not force you to load the entire data set into memory before selecting your parameters. This expedites setting up the program to run a large data set overnight or even over a weekend without user intervention. WizWhy will generate a rules file in rich text format so that the results are very interpretable and easy to double-check.

STRENGTHS: WizWhy is easy to use and can process a lot of data in a reasonable amount of time. The resulting predictions are more interpretable than that of a neural net. The program produces a list of unexpected rules, so you can see which combinations of variables have interdependent relationships to the dependent variable. It allows you to change the cut point of a continuous dependent variable, and you can specify a minimum number of examples needed to define a rule.

WEAKNESSES: If you have a lot of variables in your data set, a lot of clicking is required to toggle between the "quality" and "quantity" option for each variable. A keyboard shortcut for this would make things much easier. If you have a continuous variable where extreme high values would make a strong rule and extreme low values would make another strong rule, WizWhy will only find one of those two rules.

SELECTION CRITERIA: The predictive capabilities are excellent. We needed to supplement our neural nets with a tool that could work more effectively with large data sets containing weak patterns. WizWhy's rule induction methodology outperforms neural nets in this area. Additionally, it produces ample documentation on the rules it finds. The program is very user friendly.

DELIVERABLES: MEDai has used WizWhy to predict the probability of stocks moving up or down. A large percentage of the stocks with a probability >50 percent and a low error probability actually do move up. The models continue to work well on out-of- sample validation data.

VENDOR SUPPORT: Vendor support is excellent. Whenever I call with a technical question, the people at WizSoft are very quick to respond and provide very good answers.

DOCUMENTATION: WizWhy can be easily run from the documentation provided. Additionally, the manual explains the mathematics behind the method. The on- line help screens are very helpful.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access