Stock Price Prognosticator using Machine Learning Techniques

Abstract

S tock market price prediction is one of the favourite research topics under consideration for professionals from various fields like mathematics, statistics, history, finance, computer science engineering etc., as it requires a set of skills to predict variation of price of shares in a very volatile and challenging share market scenario. S hare market trading is mostly dependent on sentiments of investors and other factors like economic policies, political changes, natural disasters etc., Many theories were forwarded, mathematical and statistical applications in conjunction with probability, to simplify the complex process. After the advent of computers, it got further simplified but still challenging due to various external influential factors ruling the volatility of the market prices. Thus, AI and ML algorithms were being developed, but for only for next day using Linear Regression procedures.Our project aims to predict the prices of shares more precisely and accurately using special algorithms using RNN by improvising the back propagation, feedback routines to overcome the short-term memory loss involved in RNN thus providing efficiency in LS TM applications.Our project emphasizes how the LS TM applications perform with datasets of extreme, larger and minimal fluctuating data.

Utilizing Fundamental Analysis to Predict Stock Prices

“…Overall, LSTM was found to be the most productive model for forecasting time-series information such as stock close prices. The stock price prognosticator utilizing machine learning techniques was conducted using the dataset used for this study includes stock price data for each trading day from July 21, 2010 to September 28, 2018, covering the open, high, low, last, close prices, volume traded, and turnover for each stock [12]. Three machine learning algorithms were implemented: Random Forest, Gradient Boosting and Support Vector Regression.…”