PEIHAO SONG

Built a BERT model in Python and used it to pre-train inter-sentence relation features for user information based on the user data set of Huitianfu Company.

• Built a GloVe model and used it to perform statistical word frequency analysis and generate word vectors for user data.
• Trained a hybrid algorithm of two models to further improve the prediction performance, and used random forest analysis in R to classify the customer data based on the predicted data.

• Maintained and updated subsidiary data using SQL DataStack, ensuring accuracy and reliability
• Implemented Python-based data cleaning processes and polynomial fitting algorithms to estimate exposure amounts, enhancing decision-making capabilities
• Used logistic regression to predict whether customers would continue subscribing to B2B e-commerce services
• Conducted thorough analysis of supplier data, identifying trends and patterns to optimize inventory management and procurement processes
• Updated the model using Random Forest, resulting in a 0.03 increase in AUC and a0.09 improvement in sensitivity, enhancing overall model accuracy

• Supported brokerage roadshows and created detailed documents, improving research team efficiency
• Forecasted real estate trends through thorough field research and report analysis, aiding strategic decisions
• Developed Python simulations for A-share market food stocks, enhancing data-driven decision-making
• Conducted comprehensive data analysis to predict real estate trends, delivering actionable insights

• Fostered cross-functional collaboration to gather diverse insights for comprehensive monthly reports, promoting knowledge sharing and continuous learning
• Participated in writing two issues of the company's monthly report, which aimed to inform readers of the latest development in financial technology
• Tracked and reported on innovations of the latest fintech technology by relying on various sources
• Drove fund marketing strategies through data-driven user behavior analysis, leading to a 25% increase in conversion rates and a 15% rise in client acquisition
• Pioneered application of BPR algorithm to Guolian Fund user logs, improving targeted advertisement delivery efficiency by 30% and increasing user engagement by 20%