David Kwan

My HKU MSc Capstone project addressing "hidden dissatisfaction" in e-commerce through advanced sentiment analysis. Analyzed extensive Amazon reviews using multiple sophisticated models including Advanced Ensemble, RoBERTa Transformer, and Feature-Based approaches with extensively engineered features. Integrated Transformer-based ABSA processing comprehensive aspect-sentiments, uncovering subtle sentiment patterns that traditional analysis misses and enabling proactive business intelligence.

Implemented Drift Diffusion Models (DDM) to decode consumer choice behavior through evidence accumulation modeling. Analyzed multiple distinct decision-making scenarios across extensive simulated trials, revealing how bias amplification, speed-accuracy tradeoffs, and pressure sensitivity influence purchasing decisions. Achieved comprehensive insights into consumer archetypes from impulsive buyers to cautious deliberators, enabling targeted marketing strategies and optimized choice architecture.

Developed a comprehensive framework comparing state-of-the-art Multi-Objective Evolutionary Algorithms (NSGA-II, NSGA-III, MOEA/D) on benchmark problems ZDT1 and DTLZ1. MOEA/D achieved superior performance with extensive Pareto-optimal solutions in tri-objective optimization, demonstrating advanced capabilities in handling conflicting objectives. Implemented hypervolume analysis and statistical performance metrics, providing insights into algorithm selection for complex optimization challenges.

Developed a specialized neural machine translation system using an extensive Buddhist-specific parallel corpus, achieving significant improvements in domain-specific translation quality. Implemented custom tokenization for CJK characters and Buddhist terminology optimization. Demonstrated substantial specificity improvement over general-purpose models, with particular strength in translating technical terms like "prajñāpāramitā" and "five aggregates" that generic translators typically misinterpret.