My name is Tianyu Du (probably you’ve already know cuz you successfully found this page). I am currently a senior undergraduate at University of Toronto, studying economics, mathematics, and computer science. My long-term research goal is to figure out whether machine learning methods can help economic studies. In particular, my research interests focus on developing data-intensive methods for economists and financial market practitioners.
A thesis on applications of machine learning techniques on crude oil forecasting using news sentiments
This thesis focuses on forecast asset market movements from financial news. Specifically, sentimental analysis tools from natural language processing (NLP) are used to generate article-level sentiment scores. Then data science techniques including SVM, CNN-RNN are deployed to create predictive models and capture the underlying inter-temporal dependencies.
Github Repository is not Public Yet : (
Final Project for STATS202: Data Mining and Analysis at Stanford University
Patient Positive and Negative Syndrome Scale (PANSS) scores of schizophrenia patients were used to test treatment effects, k-means and Gaussian mixture were used to cluster patients based on scores prior to treatment. Moreover, SVM, random forests, and boosting machines were developed to detect potential invalid assessments and forecast patients’ future psychological states.
Independent Research Project
This project compared artificial neural networks and classical models on financial time series. Specifically, fully connected and RNN with LSTM cells were used on exchange rate forecasting, which had outperformed existing ARIMA and VAR models.
See the project page for earlier projects.