Mean reversion of stock price is an observed phenomenon in finance. It is similar to the concept of Newton's law of cooling. Three models of ordinary differential equation are derived from the concept and solved by dynamic integration. The models are applied to the daily closing price of Taiwan Stock Exchange Capitalisation Weighted Stock Index. Empirical study shows that the models have good fits and forecasts although errors appear during the dynamic forecasting process. The applicability of the models and future modification are also discussed.