Stephen Yang

I am an out-going and positive person, endowed with strong learning abilities and a robust foundation in data analysis and data mining. My skill set includes data cleaning, statistical analysis, and data visualization. Through various academic courses and competitive projects, I have honed my mastery of common machine learning algorithms and deepened my understanding of Transformer model. During my undergraduate studies, I was actively involved in the Youth Volunteer Association, taking part in and organizing a range of activities within the club and broader association. This involvement sharpened my organizational and communication skills significantly. Additionally, my participation in several university projects focused on data analysis and artificial intelligence fostered a strong sense of team spirit. In my personal life, I am committed to staying physically fit and revel in engaging in outdoor sports such as basketball, soccer, and fishing, which allow me to connect with nature. I am also an avid traveler, drawn to exploring diverse cultures and landscapes. Traveling expands my horizons and enhances my understanding of the world, which is why I find it so enriching and compelling.

Reinforcement Learning-based Combinatorial Recommendation System Framework, 06/2023 - Present
Project Overview
This project addresses the scenario of recommending lists of items composed of various combinations, derived from RS4RL data. By building a Transformer-based DQN (Deep Q-Network) reinforcement learning model that utilizes user historical behavior data and item features, the system learns to recommend the most suitable item combinations, thereby enhancing user satisfaction.
Work Content
1. Utilized Python for data cleaning and feature extraction from user shopping preferences and item information within the data.
2. Developed a Transformer-based DQN reinforcement learning model using the PyTorch framework to recommend item combinations based on user and item features, and validated its accuracy.
3. Improved the model using the Demonstration-Q-learning from Demonstrations (DQfD) method to reduce the exploration time required by the reinforcement learning agent.

Sentiment Analysis-Based Recommendation System Exploration, 08/2023-10/2023
Project Overview
This project explores the effectiveness of sentiment analysis on product reviews using models like XLNet, BERT, RoBERT, and LLaMA, and how the results of the sentiment analysis can be used to recommend similar products to users.
Work Content
1. Used Python to clean product review data and divide it into datasets.
2. Fine-tuned the XLNet model and optimized its parameters, achieving approximately 95% accuracy in sentiment analysis results.
3. Developed an algorithm to convert sentiment analysis outcomes into a cosine similarity matrix incorporating user review features.

Time Series Analysis of Public Administration and Safety Employment Numbers in New Zealand，04/2023.04-05/2023

Project Overview
This project involved analyzing the raw time series changes in employment numbers in public administration and safety from Q2 2021 to Q1 2023 as reported by Statistics New Zealand. Using the R language, different ARIMA and ETS models were constructed, selected, and forecasted. The best time series model for the data was determined based on AICc and residual diagnostics.
Work Content
1. Utilized R for STL (Seasonal and Trend decomposition using Loess) and difference analysis of the data, integrating results from ACF and related statistical characteristics.
2. Explored different time series models for the data, comparing various ARIMA and ETS models through AICc and residual diagnostics.
3. Employed the selected optimal model, ARIMA(0,1,1), for forecasting and assessed its accuracy using a test dataset.

• Southwest University Undergraduate Second-Class Scholarship, 2020
• Southwest University Undergraduate Third-Class Scholarship, 2021
• Southwest University Undergraduate Third-Class Scholarship, 2022
• Chinese College Student Mathematical Modeling Competition Second Prize, 2021
• University of Auckland Transnational Education Partner High Achiever Award, 2022