Fund Manager Analysis for MOM ProductŸ Established factor system based on multi-dimensional fund manager data, collected from both internal resources (previous due diligence files and data in the data lake) and external resources (financial data from provider and web crawler). Several NLP algorithms – LSTM, CNN, BERT were used on text data for subject recognition and sentiment analysis.Ÿ Data Preprocessing – filled up missing values by EM algorithm and other statistical features. Data transformation – converted categorical and text values to numerical values. Feature selection – implemented information gain and correlation coefficient scores to filter out effective factors.Ÿ Trained model using machine learning algorithm including SVM, random forest in 6 independent degrees suggested by clustering algorithm and adjusted by professional aspects. Graded fund manager in 6 aspects and draw persona for each fund manager. Classified fund managers into 9 investment styles and generated suggestion report based on their persona.Knowledge Base Question Answer System for Investment UseŸ Implemented semantic parser for user questions, conducted knowledge graph based intention recognitionbased on a classifier, optimized data searching route between 3 databases (ES, PG, Neo4j) and designedover 20 answer formats, including text, graph, chart and reports, under specific scenes.Ÿ Pushed recommendation questions based on semantic similarities and statistical analysis of userbehaviors.Ÿ Helped users to search overall news, trading data, operation circumstances and relations betweenenterprises, people and events. Also provided comprehensive financial analyzing reports for investors injust a few seconds.

Developed a web application for data analysis and its visualization based on questionnaire dataBuilt a record management system (in Java) using SpringMVC and MyBatis framework

 Conducted background research of car rental market, analyzed competitors with their primary business, corresponding quantities, price range and operation efficiency. Engaged in data mining from several sources. Demonstrated data validity by multi-source cross verification and model deduction. Structured out car rental market components and built a model of market share based on it.

Engaged in developing a platform for automatic testing of collimator, which is a core component in CTConducted collimator’s reliability tests and designed a collimator testing interface (in Python)