2014 - Time Series Prediction for the Tweets Popularity with Multivariate INAR(1) process
This is my currect project as well as my thesis project. In this project I built a temporal model to predict user's tweets popularity in Twitter and
improved the prediction accuracy by combining the tweet topic feature.
2014 - Topic Modeling in Twitter
- Built a model to generate topics from Twitter dataset.
- Improved the quality of topic detection with a Pitman-Yor process-based model.
- Developed a stopword-based language identification method.
- Improved the language identification speed by eight times compared to contrastive approaches while obtaining 98%+ accuracy.
2013 - Community Membership and Dynamics
Network Theory course project
- Investigated community structures of social networks%: Live Journal, a social blog website, and DBLP, a computer science bibliography website.
- Discovered that online communities are often built around common interests shared by people, and authors in DBLP prefer following topics to following each other.
2013 - racking Human Faces and Eyes Affected by Scale Variations
Computer and Biological Vision course project
- Developed a face detection system to track faces and eyes in video clips by using Viola-Jone's detection model combined with Hyper-complex Fourier Transform saliency detection method.
- Capable of tracking multiple faces of different sizes and orientations at the same time.
- Improved the eyes detection algorithm performance by using a skin cone filter as a post-processing step.
2012 - SVM-based Prediction Model for User Preference in Social Network
Machine Learning course project
- Built a SVM-based prediction model to learn user's preference for automatic recommendations.
- Improved the accuracy of prediction to 74% with context and network structure information by generating features based on the context and the user's relationships with others.
2012 - Role-based Click-through Rate Prediction Model for On-line Advertising
Bachelor Degree Thesis
- Developed a prediction model of user's online preference.
- Improved the pCTR (prediction of the Click-through Rate) accuracy by examining user's preferences with cluster analysis of sparse matrix.
2010 - A Quick Grayscale-based Face Recognition System
Digital Image Processing course project
- Developed an application to detect faces in pictures with complex background using Matlab
- .Improved the recognition speed by using grayscale features to identify face boundaries.
2010 - Online Bookstore and Background Information Management System
Project Leader, Advanced Database Systems course project
- Built a e-bookstore including UI and databases.
- Designed and built databases in SQL and C++.
- Fulfilled requirement analysis with UML.
2010 - Luoyuan Blog, Campus Blog System
Testing Engineer, Microsoft Summer Intern
- Built a on-campus blog website.
- Worked as a software testing engineer. Wrote test plans. Designed test cases in C# with LoadRunner and VS Test Tools
- Created the "drift bottle" function to increase the enjoyment and expand the scope of readers at the same time.
- Won the Best Project Prize of the summer software internship.
-> Demo Vedio