Winter 2019/2020
Pacman Gaming AI
Training a Deep Q Network to play Pacman and achive as many points as possible. Can it manage to beat a human player?
Fall 2019
Natural Language Processing
People are leaving the state of Illinois. But why do they leave and how can this trend be reversed? Applying NLP to articles from the Chicago Tribune (local newspaper) I try to find answers to these questions.
Spring 2020
Real Time Trading System
Before you start trading, you need an environment. In this project, I build a fictious server and a client that exchange market data. The client creates an orderbook and keeps track of your portfolio. Three different trading strategies function as the client's brain for decision making. Ready to make some money?!
Fall 2019
ML for Anomaly detection
Ad fraud is big. In fact, about 90% of ad clicks online might be fraudulent, only to increase clicks and generate more revenue for the website or app provider. Using different machine learning techniques, I try to uncover apps or websites that try to falsly increase their click-rate.
Spring 2019
Relational Database Management
Divvy is Chicago's local bike sharing service. To assist their decision making, I create a relational database that feeds into a Tableau dashboard and informs Divvy about busy routes, rush hours by location or most interesting neighbourhoods for expansion.
Summer 2019
ML for optimizing scouting in soccer
In today's world of professional soccer, money is often seen as key to success. But how can smaller clubs manage to keep up with the big fishs? Analytics might be a way to do so. In this project, I create a machine learning pipeline for Chicago's local MLS club Chicago Fire. The pipeline is able to assist the club in finding interesting players, identify the undervalued ones and make a reasonable first offer.
Winter 2019/2020
Recommendation System in pySpark
For streaming platforms, keeping customers on board is crucial, which is why a good recommendation system is the key to success. Since there is an array of reasons why a customer might like a certain movie - like favorite actors, genre, directors - the size of data such recommendation engines need to handle grows rather quickly. In this project, I am building a hybrid recommendation engine based on IMBD and Netflix data using pySpark.