Stock Price Prediction with LSTM, Attention and Convolution

EasyChair Preprint 11418

Abstract

This project proposes a novel machine-learning method and model for predicting closing stock prices. It does so by combining many recent existing advancements in the field into a sophisticated model. The model is used on Yahoo Finance datasets to predict IBM prices, on which it sees great performance improvements over common models, and to predict S&P 500 prices, on which it sees similar performance to other models. While it is still far from being able to beat the stock market on its own, models like these are becoming smart enough to become potentially viable for short-term, small-scale investments, due to their ability to predict closing prices one day out with decent accuracy.

Keyphrases: Attention is All You Need, Convolutional Networks, LSTMs, Stock price prediction, machine learning

@booklet{EasyChair:11418,
  author    = {Aashay Chaudhari and Benjamin Middleton},
  title     = {Stock Price Prediction with LSTM, Attention and Convolution},
  howpublished = {EasyChair Preprint 11418},
  year      = {EasyChair, 2023}}