Skip to content
View moscmh's full-sized avatar

Block or report moscmh

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
moscmh/README.md

Man Ho Cheung (Mos)

Graduated in September 2025 at Curtin University with a master's degree in Predictive Analytics, in Perth, Australia. Passionate about revealing actionable insights and building solutions through data visualisation, feature engineering, and predictive modelling in a machine learning workflow and enhancing operational efficiency using AI.

Portfolio

Any feedback is appreciated.

The additional feature improved the model's performance, $R^2$, significantly to over $.96$.

  • Data visualisation (matplotlib, seaborn)
  • Feature engineering (Gaussian Mixture Model)
  • Regression (scikit-learn)

The model with an exogenous feature, temperature, estimated that the fuel price in Perth will fluctuate around ADU177/100L in the coming two years, with a relatively lower price from November 2025 to April 2026.

  • Box-Jenkins's modelling framework
  • Data manipulation (MySQL)
  • Time-series analysis (SARIMAX)

This immersive data science workflow demonstrated how crucial exploratory data analysis and data preprocessing are to the model's performance.

  • Downsampling to address extremely unbalanced classes, which improved the model's accuracy from 50%+ to 70%+.
  • Enhanced model training efficiency with a data processing pipeline (sklearn.pipeline)
  • Hyperparameter tuning using Optuna

The interactive web application aims to raise awareness in the Hong Kong precious species that are prone to extinction. The species occurrence predictions from the CNN-LSTM deep learning model also help local green groups with environmental monitoring and conservation planning. Live Demo

  • Contains 2001-2024 Hong Kong 1000+ species occurrences geospatial data
  • Data congregation and preprocessing for model training
  • CNN-LSTM model to predict approximate locations
  • AWS Q Developer UI design and AWS Lightsail instance for web application

Tools

Language

Python, R, SQL

Visualisation

Tableau, PowerBI, Python (matplotlib, seaborn)

Machine Learning

Scikit-Learn, Tensorflow, PyTorch, SARIMAX

Certification

Microsoft Azure Data Scientist Associate Certification

Job-Related

I am looking for jobs where I can apply my knowledge in a real-world setting. Apart from that, I am also interested in project collaborations.

Pinned Loading

  1. portfolio portfolio Public

    Jupyter Notebook