Projects

• This study leverages large language models (LLMs) with prompt engineering for the detection of cyberbullying in English, Chinese, and Korean. 

• The results show that zero-shot prompting consistently outperforms few-shot prompting. Claude 3 Sonnet outperforms GPT-3.5 and GPT-4 across all tested languages.

KEYWORDS: Cyberbullying, Large Language Models, Prompt Engineering

• Implemented time series forecasting models for demand values, inventory management, and stock control.

KEYWORDS: Time Series Forecasting, Inventory Management, Stock Optimization

• Implemented a map-matching algorithm to analyze telematics data, correlating each data point with the corresponding road segment within the established Geographic Information System (GIS) network.

KEYWORDS: Map-Matching, Telematics Data, Geographic Information System (GIS)

• Developed and evaluated a trading strategy to maximize returns using the Relative Strength Index.

• Achieved optimal return of 432.52% compared to the baseline return of 74.6%.

KEYWORDS: Linear Programming, Relative Strength Index, S&P 500

• Utilized machine learning models and graph neural network models to analyze co-purchasing behavior on Amazon.

• Graph neural network models demonstrated superior performance, achieving an AUC of 0.84 compared to other models.

KEYWORDS: Recommendation System, Graph Neural Networks, Co-Purchasing Behavior

• Employed multiple clustering algorithms to analyze cities, identifying K-Means as the optimal algorithm based on Silhouette Score for insightful demographic-based comparisons.

KEYWORDS: Urban Depopulation, Machine Learning, Clustering

• Produced a comprehensive report on the company's online marketing performance.

• Utilized Twitter API to gather tweets mentioning the company and conducted sentiment analysis.

KEYWORDS: Natural Language Processing, Sentiment Analysis, Online Marketing

• Conducted sentiment analysis on the collected tweets using NLTK and Textblob, providing insights into public opinion surrounding the protests.

• Created a word cloud based on the frequency of words mentioned in the tweets, offering a visual representation of the most common topics associated with the protests.

KEYWORDS: Natural Language Processing, Sentiment Analysis, Word Cloud

• Utilized dataset provided by New Taipei City Government.

• Created data visualizations to present the distribution of YouBike stations in New Taipei City, Taiwan.

KEYWORDS: Data Visualization, YouBike, Open Data