Predicting the price of a stock via regression model I am trying to predict the closing price of a stock on a given day given opening price, the highest value and lowest value for that day. I just wanted to check if this model sounds alright:
$$ClosingPrice_{t}=\alpha_1 \cdot OpeningPrice_{t}+\alpha_2 \cdot HighPrice_{t}+\alpha_3 \cdot LowPrice_{t}+\varepsilon_t$$ 
I plan on using ridge regression as I suspect there will be multicollinearity between some of these variables.
Any feedback or thoughts on my model would be great.
 A: Let me denote the closing price $C$, the opening price $O$, the high price $H$ and the low price $L$.
Some thoughts:


*

*This is not an ARIMA model (as stated in the original title), this is a regression model. 

*This will not be a feasible predictive model since $H$ and $L$ are not available until the trading closes, when also $C$ becomes available. 

*You may introduce some coefficient restrictions to ensure the closing price always is inbetween $H$ and $L$ (possibly equal to one of them). 

*Excluding the intercept as you have done may or may not be a good idea. Including it could make the interpretation more straightforward. Also, for the sake of interpretation you could consider using $H-O$ and $L-O$ as regressors in place of $H$ and $L$, respectively. 

*Stock prices are nonstationary, integrated processes. Running a simple regression with a bunch of integrated variables will yield OLS coefficients with nonstandard properties. Since $C$, $O$, $H$ and $L$ will almost certainly be cointegrated, a vector error correction model (VECM) seems to be the relevant model. It also has the virtue of being suited for forecasting. If you apply it on logarithmic data, you will have logarithmic returns on the left hand side, which is quite convenient.

