Marketpulse - ML-based Stock Market Trend Analysis Application

Personally contributed on training a deep neural network using Tensorflow Keras, on analysing stock-based tweets to derive sentiment of the future trend of a stock. Additionally, worked on creating the admin-side interface, where new models can be uploaded, performance graphs and confusion matrices can be viewed, and existing models up on the application can be used as starting points to train new versions. Lastly, analysed the explainability of the stock tweet-analysis model, using LIME as an output-significance analysis tool for text-based classifiers.

A collaborative project done in the course “Software Engineering for Data-Intensive AI Applications - DIT826” at the University of Gothenburg. (view repo)