The Application of Neural Networks to Stock Pricing
Our equity-pricing model uses neural networks to learn how a stock’s fundamental balance sheet data has been translated to a market price in the past, within the context of its industry group and economic sector. Common factors such as earnings, sales, margins, dividends, debt, book value, and cash flow are used to train our networks. A separate network has been created for each economic sector, and is used to estimate prices going forward.
How Are Stock Prices Determined?
The latest quarter's information is fed into the neural net models to estimate current fair market prices for any stock. Our processing is carried out once a month and the results are delivered as a spreadsheet or a database for use with TradeStation. These values can then be used in many ways.
Price Disparity is Predictive
Any disparity between estimated prices and market prices is predictive of average future price trends six to twelve months in the future, and so has a multitude of uses. Price Wizard prices lead to improved stock screening, aggregate estimates of ETF value, estimates of asset class value, risk management, and the building internal index portfolios. During the 2000 stock market bubble, our model helped avoid being seduced by the herd (see picture below).
Each month Parallax aggregates Price Wizard prices by sector, industry, market cap, and asset style. We also estimate the median valuation of all US equities. When the median valuation is beyond 5%, the entire market usually responds over the next three months. Below are some recent aggregate valuation results: