Inhalt des Dokuments
Es gibt keine deutsche Übersetzung dieser Webseite.
Master Thesis: A hybrid approach to stock market predictive analysis based on microservices architecture
A hybrid approach to stock market predictive analysis based
on microservices architecture
In the financial sector, anticipating prices is a challenging task for researchers because of its noisy, nonlinear, non-stationary and dynamic nature. Forecasting stock market prices includes historical data, technical indicators, news, social media sentiments and psychological factors. In this thesis, we developed machine learning-based approach that utilises historical data and twitter data in order to implement two-week forecasting for ten stock market prices. Long short term memory model (LSTM) was used as a machine learning approach. To investigate the effect of twitter data and to determine its sentiment score, a dictionary-based approach known as Valence Aware Dictionary and sEntiment Reasoner (VADER) was wielded. The research was carried on the stock data of the organizations in technological sector such as Spotify, Microsoft and so forth. Furthermore, with the amelioration of software development, software industry is shifting to the paltry applications structures i.e. microservice architecture from a gigantic application structures i.e. monolith architecture. Moreover, we implemented a microservice-based architecture, to yield a system for machine learning techniques to be executed in a distributed system. Theil value of 0.00714 was achieved which depicted model accuracy indicating repercussion of news in stock market.
Supervisor: Hai Dinh Tuan , Maria Mora Martinez , Katerina Katsarou 
Type: Master Thesis
Duration: 6 months
10587 Berlin, Germany
Phone: +49 30 8353 58811
Fax: +49 30 8353 58409