Twitter Sentiment Classification incorporating user and event information
- Propose a comprehensive neural network model by incorporating user information and underlying event information into document-level Twitter sentiment classification using Deep learning neural networks like CNN, RNN, LSTM, GRU.
- Multiple experiments have been implemented to verify the performance of sentiment classification on a large-scale Twitter dataset using Keras tools on Theano.
Modeling the March Madness March
- Learn which learning models, data processing technique, ordering technique and parameters tuning yielded the best classification accuracy.
- Using machine learning algorithms like SVM, Naive Bayes, Random Forest, Neural Network, XGboost, etc.
IMDb Database Exploration
- Created and explored networks from IMDb movie data
- Studied the properties of actor/actress graph and movies graph using PageRank, Community Finding- FastGreedy Newman algorithm, etc., and predicted movie ratings based on regression model and graph model.
Human Attributes Extraction in Profiles March
- Collect many raw profiles in websites by our configurable crawler.
- Use statistic methods to do pre-processing works include lexical analysis like Tokenization, POS tagging, parsing, and name entity recognition using method of linear chain Conditional Random Field (CRF), and Entities relation extraction.
- Build the semantic patterns, which can be used in model to predict the person attributes.
Recommendation System using the Movie Lens dataset
- Predict the rating of all the movies using Collaborative Filtering, alternating least squares and 10 fold cross validation.Once the ratings are predicted they can be used to recommend top N movies to each user.Popularity Prediction on Twitter January 2016
- Predicted which topics will become popular on Twitter using Linear Regression and Super Bowl hashtags.