April, 2021 – March 1, 2024• Constructed the RSRS market timing model and made some changes to improve the annualized rate of return.• Built different strategies frameworks with CNN, LSTM, and Transformer based on price volume factors, and analyzed their performances• Built different shares selection frameworks using such as random forest, SVM, XgBoost and compared their performances• Performed single factor analysis for A-share stocks according to the market performance and… Show more April, 2021 – March 1, 2024• Constructed the RSRS market timing model and made some changes to improve the annualized rate of return.• Built different strategies frameworks with CNN, LSTM, and Transformer based on price volume factors, and analyzed their performances• Built different shares selection frameworks using such as random forest, SVM, XgBoost and compared their performances• Performed single factor analysis for A-share stocks according to the market performance and daily research needs• Built a multi-factor framework for Futures to follow the market trend• Constructed Fund Portfolio Performance Analysis, to perform the Net Asset Value (NAV) calibration, Profit and Loss (PNL) reconciliation & attribution analysis, and correlated risk statistics computation• Constructed portfolio of different assets to satisfy different investors’ customization needs• Constructed fund performance assessment modules with “Datayes” and built the customized platform of invests and research.• Performed due-diligence researches with different machine learning framework strategies or techniques Show less

Designed clustering algorithm based on the logic of Hierarchical Clustering and frequency statistics for all the financial statements data of more than 1000 tickers• Utilized Python (Pandas) and Bloomberg API for data analysis projects for new functions and database maintenance with large data sets (BQuant, Excel worksheets), including data cleaning and clustering, model building and statistical analysis• Conducted factor analysis to decide segment importance for more than 1000… Show more Designed clustering algorithm based on the logic of Hierarchical Clustering and frequency statistics for all the financial statements data of more than 1000 tickers• Utilized Python (Pandas) and Bloomberg API for data analysis projects for new functions and database maintenance with large data sets (BQuant, Excel worksheets), including data cleaning and clustering, model building and statistical analysis• Conducted factor analysis to decide segment importance for more than 1000 tickers• Concluded consensus through investigating financial statement forecasts and estimates at least 3 companies from the top 2000 companies in the world to ensure all the brokers’ estimates are on the same accounting basis for industrial companies including banks, technology companies, and chemical companies Show less