Our complete project report is available in the TU Delft repository.

As bachelor end project, I built a platform for sentiment analysis of social media together with four other students. The project was commissioned by the TU Delft Faculty of Architecture and the Built Environment with the goal to gain novel insights in urban planning.

We implemented a data pipeline with Apache Kafka to stream-process geotagged tweets within the Netherlands and Flanders. The first step was using machine learning for sentiment analysis, in order to categorise tweets as positive, negative, or neutral. Next, to add some demographics to the data for research purposes, we used facial recognition models to estimate the age bracket and gender of a user based on their profile picture (including a disclaimer regarding the reliability of this data).

To interact with the data and make it a useful product for researchers, we created an interactive dashboard with React, backed by a Python back-end that served a REST API using Flask. In the app users could easily query and visualise data and make comparisons between different segments.