Sheng Wai

Sheng Wai

NTU Applied Math

Projects

In reverse chronological order based on the ending date:

  • Stock Options Pricing in Financial Markets with Machine Learning | [Code]
    • Focus: Financial Engineering, Quantitative Finance, Machine Learning
    • Developed a machine learning model for stock options pricing using a dataset of 74,492 rows and 21 columns from Yahoo Finance, covering 50 companies across multiple sectors.
    • Conducted exploratory data analysis, feature engineering, and model optimisation to identify key factors influencing option pricing, such as bid price, stock price, and moneyness.
    • Implemented and evaluated multiple regression algorithms (Linear Regression, Ridge Regression, Random Forest, Gradient Boosting, Support Vector Regression), selecting Random Forest as the best model with a 95.66% R-squared value after hyperparameter tuning using RandomizedSearchCV.
    • Visualised model performance and prediction errors through scatter plots, histograms, and feature importance graphs to analyse pricing accuracy and identify critical factors driving pricing decisions.
    • Tech Stack: Python, Pandas, NumPy, Statsmodels, SciPy, scikit-learn, Matplotlib, Seaborn


  • Dynamic Programming Heuristics for Solving the Quadratic Knapsack Problem (QKP) | [Code]
    • Focus: Algorithm Design, Computational Complexity, Operations Research
    • Implemented dynamic programming heuristic algorithms from scratch to solve the Quadratic Knapsack Problem (QKP), based on the pseudocode outlined in a research paper.
    • Designed a testing framework to validate the algorithm’s correctness and provide a detailed complexity analysis of the proposed algorithmic approach.
    • Developed test cases to conduct performance evaluations, ensuring both the optimality and computational feasibility of the solutions in alignment with theoretical expectations.
    • Tech Stack: Python, NumPy


  • Heart Disease Prediction Using Machine Learning | [Code]
    • Focus: Machine Learning, Data Science, Data Analysis
    • Built an ML pipeline including data preprocessing (handling missing values, outliers), exploratory data analysis (EDA), and feature selection using Mutual Information and Chi-squared tests.
    • Conducted comprehensive model evaluation to ensure robustness and data quality, optimising for key metrics such as accuracy, precision, recall, and F1 score.
    • Enhanced model performance through hyperparameter tuning using GridSearchCV for K-nearest Neighbor, Logistic Regression, Random Forest, and Multi-layer Perceptron.
    • Achieved a 16.7% improvement in recall and a 16.3% boost in accuracy for Random Forest, making it the best-performing model among the four tested.
    • Tech Stack: Python, Pandas, NumPy, scikit-learn, Matplotlib


  • Quantron | [Code] [Live Demo]
    • Focus: Full-Stack Web Development, Chatbot Development, Natural Language Processing
    • Developed a full-stack chatbot web application using Flask that entertains users through text-based communication.
    • Utilised NLTK for rule-based text generation, enabling the chatbot to understand and respond to a wide range of user queries with predefined patterns and reflections.
    • Integrated external APIs via HTTP requests, including Google Translate for translation services and a joke API, enhancing user engagement and providing diverse functionalities.
    • Implemented full-stack integration by exposing backend services via RESTful API endpoints using Flask and Fetch API in JavaScript, enabling real-time communication between the chatbot system and the frontend.
    • Tech Stack: Python, HTML, CSS, JavaScript, Flask, NLTK


  • Quantum Key Distribution (QKD) Simulator | [Code]
    • Focus: Quantum Cryptography, Information Security
    • Implemented a QKD simulator using Qiskit, providing an environment for secure key distribution through quantum communication for cryptography.
    • Developed an intuitive Graphical User Interface (GUI) using Tkinter, allowing users to interact with the simulator for key generation, encryption, and decryption.
    • Created a quantum circuit with specified qubit lengths, incorporating essential quantum gates (Hadamard, X, Y, Z) to simulate key preparation, communication, and measurement processes.
    • Implemented security features, including error correction using a repetition code, eavesdropping detection, and privacy amplification through a hash function, ensuring robust quantum key exchange.
    • Tech Stack: Python, Qiskit, Tkinter, hashlib


  • BallBox Odyssey | [Code][Live Demo]
    • Focus: Frontend Web Development, Game Development
    • Developed an interactive web-based game where players control a ball to collect food while avoiding mines.
    • Designed gaming rules and algorithms, including player movement, collision detection, scoring, game-over conditions, randomised mine placement, and food item generation, ensuring smooth and engaging gameplay.
    • Enhanced user experience with a responsive UI, incorporating score tracking, live tracking, real-time feedback for food and mine interactions, and a game reset functionality for continuous play.
    • Integrated CSS animations for dynamic elements, such as pulsating mines and food items, and implemented event-driven programming to handle players’ movement and game state changes effectively.
    • Tech Stack: HTML, CSS, JavaScript


  • ChaosVision | [Code][Live Demo]
    • Focus: Frontend Web Development, UI/UX Design
    • Built a web application that generates artistic patterns inspired by chaos, allowing users to create images based on numerical inputs or mathematical expressions.
    • Developed a responsive UI/UX with live updates to shape properties and animation settings, employing CSS keyframe animations and JavaScript manipulation to achieve visually engaging effects.
    • Designed a comprehensive input validation system that supports both numerical and expression-based entries, using regular expressions and custom logic to prevent performance degradation.
    • Enhanced functionality by integrating html2canvas for client-side rendering of HTML into downloadable images and provided adjustable controls for opacity, size, and box-shadow to enhance customisation options.
    • Tech Stack: HTML, CSS, JavaScript


  • Neural Network Image Generator Web Tool | [Code]
    • Focus: Backend Web Development, Neural Network Visualisation
    • Spearheaded backend development of a neural network generator web application, enabling customisable 3D visualisations of neural architectures through user inputs.
    • Designed algorithms for precise node positioning, dynamic node colouring based on layer types, and automated size and line width adjustments based on neuron count, enhancing the visual interpretability of neural models.
    • Developed a Flask-based framework for real-time generation of neural network diagrams, rendering them into 3D plots with Matplotlib for display on the frontend.
    • Collaborated with project members and aligned development efforts to deliver the project.
    • Tech Stack: Python, Flask, Matplotlib, NumPy