Time Series Analysis of Stock Price Prediction Using Hybrid Deep Learning Neural Network

Abstract

The motive behind researching on “TIME SERIES ANALYSIS OF STOCK PRICE PREDICTION USING HYBRID DEEP LEARNING NEURAL NETWORK” is to explore the thoroughness of the historical financial data of a stock suffice to make cabbalistic foreboding about its aspect prices with the use of Machine Learning. The purpose of my task is that, as the price of a stock fluctuates with time dimension, it is occupied to go through with specific patterns which I prospect to capture using Deep Learning and utilize for future predictions. At first, I will amplify on the necessary of theoretical background information regarding Machine Learning , concentrating on particularly the neural networks that will later be used. Pursuing that, I will experiment that how existing research about stock market forecasting using corresponding techniques prosecuted in the past and I will propose a model in this research using Hybrid Long short-term memory (LSTM), a Recurrent Neural Network architecture which is the most suitable method for this kind of analytical tasks. I have worked with around 1132 data which were collected from online platform. Here, I have analyzed error fluctuation of the collected data through one proposed model rather than analyzing the result of different model accuracy. At last, I will try to make predictions about the future trajectories of the stocks’ prices and draw consequences from them.