Predicting intraday prices in stock market transactions using similarity profiled temporal associations

The primary intension of any investor in the stock market is to catch the market trends at an early stage and accordingly transact (buy or sell) at the right time. Though stock market data is convertible into some form of multiple time series, it is difficult to process, analyse and mine manually. Researchers have proposed several methods to predict the future price of the stocks. In this paper, we proposed a method to predict the intraday price of a stock using the historic data. Given the time stamped transactions, the stock data is mined for pattern records using similarity profiled temporal association mining with reference to a cut-off value and for forming a pattern database. Using the support value for different price gain and the opening price of the stock for the day, we extract all the significant pattern records from the pattern database. Using the current trend of the stock, we project the future prices from time to time for the day. Wipro stock data from 2005 to 2009 are used for experimental evaluation of our approach. Expected price for various days are agreed to an extent of 98% with actual transaction prices.