项目

Built a reusable crawler for dynamic pages using Selenium + BeautifulSoup, supporting infinite scroll and complex DOM structures. Designed a standardized ETL pipeline to clean/parse data and load into MySQL/PostgreSQL. Created a configurable extraction framework (XPath/CSS selector/regex). Implemented error handling and retry mechanisms. Developed a Streamlit UI for task configuration, batch runs, monitoring, and export.

Automated data ingestion and storage with PostgreSQL, including scraping, deduplication, and batch inserts. Built a resilient pipeline with scheduling and interruption control for 24/7 updates. Performed feature engineering + multi-model forecasting (Random Forest, Gradient Boosting, ARIMA, Prophet). Delivered 7-day forecasts, trend indicators, and risk intervals to support buy/sell/hold decisions.

Built KPI system (UV/PV, funnel conversion, retention, repurchase) to analyze user behavior end-to-end. Created cohort datasets with consistent metric definitions. Segmented users via RFM + K-means, identifying high-value and churn-risk groups. Produced targeted marketing and ad optimization recommendations based on segment performance.

Integrated multi-system data to build a product selection metrics framework (sales volume, revenue, growth rate, conversion, refund rate, etc.). Delivered daily operations briefs for rapid decision-making. Focused on product/category performance gaps, identified anomalies, and supported business strategy adjustments.

Web-based interactive asteroid impact simulation platform using NASA open data. Led the Prediction module: API integration and visualization linkage. Refactored from 3D globe to 2D map display for better system integration stability.

Built an analysis dataset from survey data; performed cleaning, recoding, and missing-value handling. Used descriptive stats, group comparisons, and correlation analysis to explore links between social network usage and outcomes. Delivered visuals and insights for team report and class presentation.