Here are some projects that I have done.
Personal Project
Credit Card Fraud Detection
Created a classification model to detect fraud transactions with accuracy of 97% and F1 score of 80%.
Worked imbalanced dataset with 0.72% of fraud transactions by using XGBoost and SMOTE. code
Digit Recognizer
Created a classic MNIST model with 5 convolution neural network with accuracy of 99.7%. code
Built a web demo to allow client to predict their handwritten digit and uploaded image of numerical digit using Tensorflow.js and Javascript. code
NLP Tweet Disaster Detection
Created a NLP model to detect tweets with disaster alerts.
Implemented data preprocessing, data cleaning, Classifier methods, Logistics Regression with Grid Search CV to find the best parameters.
Implemented BERT model using TFHub with accuracy of 82.8%. code
To Do List
Created a to-do list web app, inspired by Microsoft To Do. The web app allows user to add new tasks, mark as finished and also remove tasks added. code
Calculator app
Created a basic calculator app, inspired by Google Calculator. code
Flappy Bird
Created a 2D demo of Flappy Bird with p5.js code
Group Project
Web Whiteboard - 3rd place
This is our Code Jam 2020 Hackathon project with Diego Lopez and Khoi Nguyen.
Created a virtual whiteboard to reduce the challenge for students to attend online schooling with Javascript, HTML, CSS and p5.js.
The webiste worked well with both mouse and touch pen. code
Modified MNIST - 1st place
Worked in a team with Ricky Chen, Peter Guanhua Rong for MAIS202 Kaggle Compeition.
Created algorithm with 5 layers convolution neural network to identify the largest numerical value in pictures that contains 3 handwritten numerical digits.
Implemented Image Data Generator using Keras library with some data preprocessing methods which allowed us to reach the accuracy of 98%. code
Hackathon
C the spped of light.
This is our project submitted to Hack McGill 8, the project is an interactive visualisation of lorentzian transforms in special relativity.