STOCK PRICE PREDICTION USING DISTRIBUTIVE COMPUTING

Abstract

The stock market, equity market or share market is the aggregation of buyers and sellers (a loose network of economic transactions , not a physical facility or a discrete entity) of stocks (also called shares); these may include securities listed on a stock exchange as well as those only traded privately. In today’s world Stock Market is a great factor in determining the state of the economy of a country. The movement of share market determines the movement of economy. Hence everyone tries to determine the movement of market on a particular day. Here we present a methodology of finding the important stocks for day to day traders with reduced time and considerable accuracy. The data sets being used are a subset of published stock data by the Bombay Stock Exchange(BSE) and National Stock Exchange(NSE) in the past one year. We collected the sentiment score of a stock. Sentiment score is a score which gives us an idea about how people feel about buying a particular stock. We used www.moneycontrol.com for finding the sentiment score associated with a particular stock, which was further used for training the neural network model. The neural network would work as a non-linear binary classifier that would predict the progress of the company using the data for the present day. Due to the large amount of data required and quicker real-time feedbacks to be supplied, we will be using Apache Spark. Apache Spark is an open source cluster computing framework for implementing distributed computing. In this framework a bigger job is distributed among a group of workers so as to reduce the actual time required for the job. With the help of Apache Spark and neural network, we would be able to find a list of companies listed on the stock market that have the tendency to go up in the day in a very short period of time.