Advanced

Predicting Tesla Stock Return Using Twitter Data - An Intraday View on the Relation between Twitter Dimensions and the Tesla Stock Return

Edman, Gustav LU and Weishaupt, Martin LU (2020) NEKP01 20201
Department of Economics
Abstract
In this thesis, Twitter data is used to predict the intraday stock return for Tesla, Inc. We present two different methods to extract the tweets’ sentiment: A dictionary-based approach (VADER) and a machine learning approach (SVM).
Additionally, we control for other dimensions as the user and discussion dimension. Then a Granger causality test and a lasso regression are conducted on a one- and five-minute interval. The results suggest that there is no predictive power in the information of the tweets for the dictionary data set and the machine learning data set. Using a subset of the dictionary data set with only the cashtag does not alter the results. The reason for this may be that we employ two linear models on a
possible non-linear... (More)
In this thesis, Twitter data is used to predict the intraday stock return for Tesla, Inc. We present two different methods to extract the tweets’ sentiment: A dictionary-based approach (VADER) and a machine learning approach (SVM).
Additionally, we control for other dimensions as the user and discussion dimension. Then a Granger causality test and a lasso regression are conducted on a one- and five-minute interval. The results suggest that there is no predictive power in the information of the tweets for the dictionary data set and the machine learning data set. Using a subset of the dictionary data set with only the cashtag does not alter the results. The reason for this may be that we employ two linear models on a
possible non-linear problem. (Less)
Please use this url to cite or link to this publication:
author
Edman, Gustav LU and Weishaupt, Martin LU
supervisor
organization
course
NEKP01 20201
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Sentiment, Tesla, Twitter, Support Vector Machine, VADER.
language
English
id
9025587
date added to LUP
2020-08-29 10:39:12
date last changed
2020-08-29 10:39:12
@misc{9025587,
  abstract     = {In this thesis, Twitter data is used to predict the intraday stock return for Tesla, Inc. We present two different methods to extract the tweets’ sentiment: A dictionary-based approach (VADER) and a machine learning approach (SVM).
Additionally, we control for other dimensions as the user and discussion dimension. Then a Granger causality test and a lasso regression are conducted on a one- and five-minute interval. The results suggest that there is no predictive power in the information of the tweets for the dictionary data set and the machine learning data set. Using a subset of the dictionary data set with only the cashtag does not alter the results. The reason for this may be that we employ two linear models on a
possible non-linear problem.},
  author       = {Edman, Gustav and Weishaupt, Martin},
  keyword      = {Sentiment,Tesla,Twitter,Support Vector Machine,VADER.},
  language     = {eng},
  note         = {Student Paper},
  title        = {Predicting Tesla Stock Return Using Twitter Data - An Intraday View on the Relation between Twitter Dimensions and the Tesla Stock Return},
  year         = {2020},
}