Predicting anything is very hard where the relationship between inputs and outputs are non-linear in nature. The prediction of stock market values is one of a challenging task of financial time series. Online application for buying and selling the shares is used in high amounts these days. The next step of this web application will be not just registering, buying and selling the shares but it will also be predicting the values for particular shares in the market. We are proposing the system which will study the database of shares and will give predictions according to it. With the help of study of neural networks the system will be designed and based on. For prediction particularly ARMA (autoregressive-movingaverage) algorithm is used. Hence the system will be able to give highest probability predictions for particular shares.