Markets are made by, and respond to, the thoughts and feelings of
human beings, including commentators and analysts as well as buyers
and sellers. Traditional "black box" and other computerized market
analysis tools can assess technical indicators, but cannot easily
integrate human sentiments without human aid.
Using Webmind's text-analysis capability, along with unique and
advanced statistical-pattern-recognition-based numerical forecasting
techniques, MarketPredictor extracts valuable information from
market-related news and opinion texts, adding
real-world
relevance and dimensionality to the traditional
business metrics used in market modeling and forecasting. The result
is a prediction tool that recognizes and interprets the significance
of events, facts and opinions as well as numbers, substantially (and
provably) increasing the accuracy of its judgments, and minimizing
investment risk.
Instead of applying one-size-fits-all algorithms and data sets,
MarketPredictor examines the reams of relevant market data that are
generated every day -- from news bulletins to analyst ratings - and
determines their relevance, trustworthiness, and applicability to
its
analysis of specific market activity. In other words,
MarketPredictor understands 100-day moving averages and peak/trough
analysis, but it can also understand the impact of a bad S&P
rating, or the effect of the price of wheat on a baked-goods
manufacturer's stock.
Having derived these market factors, MarketPredictor can
integrate them with its numerical prediction techniques to create
prediction rules that are of superior relevance to traditional
technical indicators. And because MarketPredictor has inherited
WebMind's learning capabilities, each wave of market data
strengthens its understanding and ability to predict.
While beta tests are being conducted by analysts now,
MarketPredictor's prototype has performed admirably in third-party
tests, predicting movements of the Dow Jones Industrial Average, the
S&P 500, and Eurodollar futures with superior accuracy to
non-text-based analysis.