The Kaggle Project

We are a team of 8 students who compete worldwide to develop state-of-the-art machine learning models for real-world problems. Specifically, the project competes in Kaggle competitions against thousands of teams, many of which are composed of professional machine learning scientists.


This semester, we worked on a variety of Kaggle challenges. We predicted errors in Zillow's housing prices model, built computer vision algorithms on satellite imagery, used driver data to predict which drivers would file auto insurance claims, and forecasted supermarket sales.


Our team this semester is especially competitive. We achieved 42nd place out of ~4000 submissions for the Zillow Zestimate competition and may likely be accepted for the second private round. We're also currently ranked ~100 out of ~1700 for the Ship/Iceberg classification problem.


To learn more, visit our GitHub here.

Current techniques

We are spending a significant amount of time looking into deep learning models. For example, the Ship/Iceberg challenge calls for the use of a deep convolutional model to handle the image data. We are also looking into time series analysis with neural network models, for example with LSTMs.