Academic Projects

• Context-aware Multi-stakeholder Recommender Systems: Developed an online algorithm (using Multi-armed Bandits) that recommends the most relative items to users considering the contextual features of users, relevance, and priority of each involved stakeholder to provide an acceptable level of satisfaction for all stakeholders in long term (in Python).
• Two-sided Recommender Systems: Implemented a two-sided recommendation algorithm which recommends items, provided by suppliers, to users while considering the preferences and satisfaction of both sides of the marketplace (in Python).
• Predicting Short-term Passenger Flow: Developed a hybrid Neural Network to predict short-term passenger flow at bus stops. This model extracts the spatial and temporal data patterns using a combination of Graph Convolutional model, Convolutional Neural Network, and Long Short-Term Memory (in Python).
• Predicting Bus Fullness Levels: Built machine learning models such as Negative Binomial Regression, Logistic Regression, Random Forest, and Neural Networks to predict bus load/crowding level for a bus route arriving to a bus stop within a 15-minute time interval. The accuracy of models was evaluated using a large dataset (over 100 million rows) after cleaning, data transformation and feature selection (in Python and Apache Spark).
• Analysis of Online User Behavior for Art and Culture Events: Extracted topics from tweets about a cultural event using topic modeling (LDA), applied clustering algorithms such as K-means to analysis the user behavior and developed a Decision Tree classification model to predict the user interest for the future events (using R and Clarifai API).