Skip to content

Stock Prediction using LSTM, Linear Regression, ARIMA and GARCH models. Hyperparameter Optimization using Optuna framework for LSTM variants.

Notifications You must be signed in to change notification settings

SudipBishwakarma/MSc-Dissertation-2021

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MSc Dissertation 2021

This repository contains the research and implementation for my MSc Dissertation. This work was part of a paper submitted to and published in IEEE, titled "Automated Hyperparameter Optimization in Machine Learning for Stock Prediction". I am the primary author of this work.

Link to Paper (IEEE Xplore)

Abstract

Stock prediction is the key area of focus in financial analysis. The growing amount of data and readily available machine learning algorithms has surged the amount of research in this field. This research in particular, involved in stock prediction of NEPSE using machine learning algorithms such as Linear Regression and LSTM. The research also studied traditional financial models such as ARIMA and GARCH. The analysis involved in manual and automated hyperparameter optimization via Optuna framework for single and stacked LSTM models. Initially, traditional financial models performed better than manually optimized LSTM variants. But the automated hyperparameter tuning approach significantly lowered the error loss and the single LSTM model best predicted the stock price with 7.21 RMSE score.

Overview

Stock data of NEPSE from the year 2012 - 2020 was analysed. A comparative analysis of traditional financial models such as ARIMA and GARCH, and machine learning algorithms such as Linear Regression and LSTM models were studied. Hyperparameter optimization using Optuna framework for LSTM variants was the key focus of this research. Hyperparameters such as optimizers, LSTM hidden units, dropout rates, epochs, batch size and learning rate in combination with different LSTM architectures were tested.

The files are organized as:

Citation

If you find this research useful, please cite my paper:

@INPROCEEDINGS{10079816,
author={Bishwakarma, Sudip Tiwari and Sharma, Gajendra},
booktitle={2022 Second International Conference on Next Generation Intelligent Systems (ICNGIS)},
title={Automated Hyperparameter Optimization in Machine Learning for Stock Prediction},
year={2022},
volume={},
number={},
pages={1-6},
keywords={Machine learning algorithms;Linear regression;Stochastic processes;Manuals;Predictive models;Hyperparameter optimization;Data models;Machine Learning;NEPSE;Stock Prediction;Optuna;Hyperparameter Tuning;LSTM;Linear Regression;ARIMA;GARCH},
doi={10.1109/ICNGIS54955.2022.10079816}}

About

Stock Prediction using LSTM, Linear Regression, ARIMA and GARCH models. Hyperparameter Optimization using Optuna framework for LSTM variants.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published