Efficiency of Machine Learning Cloud-Based Services vs Traditional Methods in Stock Prices Prediction

Staravoitau, Siarhei (2022) Efficiency of Machine Learning Cloud-Based Services vs Traditional Methods in Stock Prices Prediction. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (4MB) | Preview
Preview	PDF (Configuration manual) Download (2MB) | Preview

Abstract

One of the challenges while running ML jobs is IT compute resource usage optimisation. Scaling of computing resources and distributed calculations are in the mainstream of cloud computing development to optimize business IT resources and costs. This research evaluates the efficiency of cloud-based platforms, tools and features (PySpark scalability and distributed Tesnsorflow MirroredStrategy, TPUStrategy and Elephas RDD model training) over cloud-based platforms while performing Data Analytics task like LSTM Stock Price Prediction using sentiment analysis based on Twitter microblog messages. The RMSE (as well as MAE, MAPE, etc.) accuracy and model training time are used as benchmarks for the evaluation of cloud computing environment parameters. Changing model training hyperparameters (number of neurons, number of epochs, batch size) is used to change research tests’ workload using Pandas and PySpark code implementation in Google Colab Pro+ GPU and TPU, AWS EMR, AWS Lambda and Local PC with GPU environments. Best RMSE accuracy and model training times were demonstrated by both PySpark-based and Pandas-based code executed using GPU in non-distributed model training mode. Distributed model training improves model training time, but model accuracy is getting reduced. Using the AWS EMR and AWS Lambda experiments along with the Google Colab-based Keras Tuner tool with Random search hyperparameters optimization didn’t produce the expected RMSE accuracy and model training times.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning H Social Sciences > HG Finance > Investment > Stock Exchange
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Tamara Malone
Date Deposited:	06 Dec 2022 15:27
Last Modified:	08 Mar 2023 15:25
URI:	https://norma.ncirl.ie/id/eprint/5972

Actions (login required)

View Item