Advanced

Neural Networks and the Stock Market

Åström, Lars LU (2018) NEKH02 20181
Department of Economics
Abstract
In the 19th century, gold diggers emigrated from Europe to North America with the hopes of a brighter future. Gold diggers today might look to the stock market with hopes of finding the key to incredible wealth. The introduction of neural networks has revolutionized data analysis and the possibility of using them to do significantly better than chance on the stock market is investigated in this report. The aim of the report is to investigate a short period after a stock drastically decreases in price with neural networks.

Firstly, neural networks are constructed and trained to estimate price changes. Thereafter an investment strategy is constructed and evaluated. The results of the investigation is that it is possible, under some... (More)
In the 19th century, gold diggers emigrated from Europe to North America with the hopes of a brighter future. Gold diggers today might look to the stock market with hopes of finding the key to incredible wealth. The introduction of neural networks has revolutionized data analysis and the possibility of using them to do significantly better than chance on the stock market is investigated in this report. The aim of the report is to investigate a short period after a stock drastically decreases in price with neural networks.

Firstly, neural networks are constructed and trained to estimate price changes. Thereafter an investment strategy is constructed and evaluated. The results of the investigation is that it is possible, under some assumptions, to do significantly better than chance in the stock market. It appears to be possible to do this for all days from the third to the ninth after a drastic crash. The predictions are significantly better than chance, however probably not large enough to compensate for transaction costs.

It appears that the gold diggers of the 21st century, operating on the stock market, have to wait a while longer. It is possible to guess significantly better than chance if the price will increase or decrease, however the difference might not be large enough to profit from the trades. The networks that are developed in this thesis are not good enough to give rise to sufficient profits on the stock market. However, maybe it is possible to do so with slight changes in the structures of the networks, and thus earn incredible profits by investing wisely on the stock market? (Less)
Please use this url to cite or link to this publication:
author
Åström, Lars LU
supervisor
organization
course
NEKH02 20181
year
type
M2 - Bachelor Degree
subject
keywords
Neural Networks, Machine Learning, Stock Market, Efficient Market Hypothesis, Artificial Intelligence
language
English
id
8944056
date added to LUP
2018-11-05 14:36:20
date last changed
2018-11-05 14:36:20
@misc{8944056,
  abstract     = {In the 19th century, gold diggers emigrated from Europe to North America with the hopes of a brighter future. Gold diggers today might look to the stock market with hopes of finding the key to incredible wealth. The introduction of neural networks has revolutionized data analysis and the possibility of using them to do significantly better than chance on the stock market is investigated in this report. The aim of the report is to investigate a short period after a stock drastically decreases in price with neural networks.

Firstly, neural networks are constructed and trained to estimate price changes. Thereafter an investment strategy is constructed and evaluated. The results of the investigation is that it is possible, under some assumptions, to do significantly better than chance in the stock market. It appears to be possible to do this for all days from the third to the ninth after a drastic crash. The predictions are significantly better than chance, however probably not large enough to compensate for transaction costs. 

It appears that the gold diggers of the 21st century, operating on the stock market, have to wait a while longer. It is possible to guess significantly better than chance if the price will increase or decrease, however the difference might not be large enough to profit from the trades. The networks that are developed in this thesis are not good enough to give rise to sufficient profits on the stock market. However, maybe it is possible to do so with slight changes in the structures of the networks, and thus earn incredible profits by investing wisely on the stock market?},
  author       = {Åström, Lars},
  keyword      = {Neural Networks,Machine Learning,Stock Market,Efficient Market Hypothesis,Artificial Intelligence},
  language     = {eng},
  note         = {Student Paper},
  title        = {Neural Networks and the Stock Market},
  year         = {2018},
}