twlottomldl

My part-time experimental project. Using machine learning and deep learning to predict lottery winning numbers.\

Kick-off: 2020/05/11
For details, please visit my project journal on HackMD: Project Journal - Taiwan Lottery Machine Learning (twlottomldl)

Snapshot

Purpose: Predict winning numbers of Taiwan lottery by ML/DL
Rationale:
- There's no perfectly fair lotto balls nor perfectly fair lotto machines existed in this world.
- The floating ,collision of balls, and (therefore) the results might be influenced by intrinsic and extrinsic factors
  - intrinsic factors:
    - characteristics of balls and machines themselves
  - extrinsic factors:
    - local weather (e.g., Atmospheric pressure, humidity, temperature, etc.) at or near the opening time.
    - gravitational forces from near by celestial bodies. (The moon should be the major source of force)
Language: Python
Techniques:
- Web scraping: requests, pyquery
- ETL : pandas(local machine)
- EDA : numpy, pandas, matplotlib, Tableau
- Modeling: (TBD)
- WebApp Deployment: (TBD, maybe Google APP engine)
Data:
- Historical lottery data
- Weather data
- Moon-phase data
Link:
- Project_Journal _twlottomldl
- GitHub_Repo_twlottomldl

Approach/ Strategy (Outline of Stages)

Stage 0: Background Survey
- Learn lottery rules
- Survey available data source of winning number & weather (& moon phases)
Stage 1: Web Scraping
- Get desired weather data
- Get desired historical lottory data
Stage 2: ETL (Extract, Transform, Load)
- Extract & correctly combine the datasets in hand.
Stage 3: EDA (Exploratory Data Analysis)
- Explore data characteristics
- pandas, matplotlib, seaborn
- Tableau
Stage 4: Modeling
- Divide Training/Testing Datasets
- Train/Test different ML/DL for better prediction
Stage 5: WebApp Deployment
- Wrap the model into a usable application
- Deploy the APP on the web