Wind Power Prediction

I developed an end-to-end machine learning project to forecast the hourly outputs of major wind farms in South Australia. I built a data pipeline in Python to ingest and clean public weather & power data. I used various models, such as XGBoost, to discover the relationship between the weather and power data. Finally, I deployed the models on AWS to continuously make forecasts based on live weather data, and built a dashboard in Dash as a frond end to demo the live forecasts from my models.

Check out the Web App!

Check out the Notebook & data in Git repository

Explainable Wind Power Forecast Workflow

I built a semi-automated data science pipeline and an explainable workflow with Lale & AIX360 to forecast wind power based on weather data. This workflow allows data scientists to better understand how model performs so they can choose hyperparameters and features accordingly to continuously re-train models.
This is my submission to IBM AI Explainability Hackathon, which won second place.

Checkout project page!

Job Posting Analysis

I used Selenium and BeautifulSoup to web scrape job postings data from au.indeed.com. Then I used natural language processing, SVM and random forest models to determine crucial factors that impact the salary.

Check out the GitHub repository

Data Visualisation with Blender & Python

I made a 3D time-series choropleth map to visualise the growth of small-scale solar PV installation in all Australia from 2001 to 2020. The 3D scene was built in Blender programmatically using solely Blender's Python API. I also built a dashboard in React.js so you can explore all the data in each postal areas.

Check out the video!
Check out an interactive demo!
Check out the GitHub repository!
Check out the dashboard!