Portfolio

---

📈 Sentiment Meets Stock

A data-driven project aims to predict SPY’s daily movement direction (S&P 500 ETF) by analyzing financial news sentiment. This end-to-end pipeline integrates text processing, sentiment analysis, and machine learning to enhance market prediction accuracy.

🛠 Tools: Python, FinBERT, OpenAI GPT API, VADER, XGBoost, ARIMA, LSTM, scikit-learn, pandas, yfinance

✨ Highlights:

• Collected and processed SPY price data and CNBC news headlines/descriptions
• Applied four sentiment analysis methods (FinBERT, GPT, VADER, LM Dictionary) and aggregated multi-source features
• Engineered rolling features and technical indicators to enrich model input
• Built and compared models (ARIMA, RF, LSTM, XGBoost); RF achieved the best performance
• Simulated trading strategy based on model predictions to evaluate potential returns
  

---

🤖 LLM-based PDF Chatbot Demo

An experimental AI chatbot designed to extract insights from PDF documents using state-of-the-art large language models. The system allows users to upload documents and interactively query their content in natural language.

🛠 Tools: Python, LangChain, OpenAI API, FAISS, Streamlit

✨ Highlights:

• Built end-to-end pipeline to process, chunk, embed, and store PDF content using LangChain and FAISS
• Implemented real-time user query interface with Streamlit
• Experimented with different prompt strategies to improve LLM response accuracy and context awareness
• Integrated multiple LLMs (ChatGPT, Claude) with modular support for switching providers

🔍 More details?：GitHub Repo 

---

🍷Iowa Liquor Sales Data Analysis

A data analysis project using public liquor sales data from the state of Iowa, with the goal of uncovering trends in consumer behavior, seasonal patterns, and store performance.

🛠Tools: Python (Scikit-Learn, pandas, matplotlib), PySpark, Tableau

✨Highlights:

• Data Integration Across Diverse Sources
• Normalized store performance based on operational years
• Optimized Machine Learning models to predict seasonal and holiday-specific trends
• Built interactive dashboards to display product-category and location insights

🔍More details?：Report | Tableau Dashboard | YouTube

---

🖥️ Intelligent text big data analysis display platform

Aimed to collect, analyze, and improve the quality of large-scale campus news text data from different sources, focusing on text search, keyword extraction, sensitive content detection, and text management.

🛠Tools：Python, NLP(TextRank, LDA), Flask, Elasticsearch, shell

✨Highlights:

• Interpreting vague business needs and translating them into feasible algorithmic solutions.
• Responsible for designing and implementing the text preprocessing pipeline using Python, including data cleaning, keyword extraction with TextRank, similar text recommendations, abnormal link warning, and sensitive word detection.
• Worked with teammates to define interface requirements for frontend-backend integration.

🔍More details: Project Summary Slides

---

🛣️ Highway Big Data Analytics and Visualization Dashboard

A highway big data analytics and visualization dashboard to assist transportation authorities in monitoring traffic flow and transaction performance across regions.

🛠 Tools：SQL, shell, dashboard tool

✨Highlights:

• Designed and implemented SQL queries to retrieve and process key vehicle and transaction data required by the client.
• Navigating unfamiliar report tools and integrating with backend services smoothly.
• Developed communication skills through continuous collaboration with teammates and coordination with stakeholders.

🔍More details:Project Website | Project Summary Slides

---