CrowdFlower Search Results Relevance Kaggle Competition
- Built machine learning model predicting search query relevance (view on Github).
- Implemented NLP algorithms for preprocessing (stop word removal, porter stemmer, TFIDF) and an ensemble of Adaboost, Random Forest, and SVC algorithms for modelling.
- Ranked 42 out of 1,326 competitors (view results on Kaggle).
Caterpillar Tube Pricing Kaggle Competition
- Built machine learning model predicting supplier quotes for tube assemblies (view on Github).
- Performed preprocessing and feature extraction using Pandas Python package and ensemble algorithms for modelling.
- Ranked 50 out of 1,323 competitors (view results on Kaggle).
- Developed web application to maintain online course list and generate resume websites (with PostgreSQL, Flask, and Bootstrap frameworks).
- Implements user accounts (Google login using OAuth 2.0), course image uploads, and RESTful API (JSON and XML).
- Built interactive data visualization of loan risks and returns using dimple.js and d3.js.
- Conducted data exploration in Tableau and preliminary data aggregation using pandas/numpy/csv Python packages.