This project involves designing neural network-based models- namely a Convolutional Neural Network model, a Long-Short Term Memory model and a Multi-Layer Perceptron model- to predict the closing stock prices of Amazon based on a daily time-series dataset. The predicted prices were then compared with the actual closing prices, to calculate the error of each neural network model. This error was further visualised using various data visualisation plots including bar plots, scatter plots and line plots, which helped demonstrate the accuracy of the model with least error, and, each plot helped extract different features of the model which can help understand if it is suited to an application based on requirements. This architecture can be further improved and expanded to improve accuracy and accommodate more applications.
Dataset Source: The data is taken from Yahoo Finance website. In this study, we have worked on the stock prices of an e-commerce company Amazon (AMZN) spanning 10 years.
You can find our presentation here- https://prezi.com/view/FwzUog9q68cnx45RfnAj/