BD528-software_engineering

ML Streamlined Workflow

MLflow

Deployment Architecture

Kubernetes

Continuous Integration (CI)

GitHub Actions

Continuous Deployment (CD)

ArgoCD

Web-UI

Diabetes Checking Web App

How to manual run? (FOR USERS)

STEP 1: Clone repository

git clone https://github.com/NutBodyslam053/BD528-software_engineering.git

STEP 2: Create conda environment & Activate it

conda create -n <env_name> python=3.9 -y
conda activate <env_name>

STEP 3: Install dependencies

pip install -r requirements.txt

STEP 4: Run application

python app.py

STEP 5: Open web browser

Diabetes Prediction Web-UI: http://localhost:5000

How to run with Docker? (FOR USERS)

STEP 1: Pull image & Run app container

docker run -dp 5000:5000 nutbodyslam053/bd528:v1

STEP 2: Open web browser

Diabetes Prediction Web-UI: http://localhost:5000

Workflows (FOR OWNER)

1. Create Github repository & Clone to local repository

git clone <github_repository_url>

2. Create template structure & Run to create needed directories

touch template.py
python template.py

3. Create virtual environment & Install dependencies

conda create -p venv python=3.9 -y
conda activate venv/
pip install -r requirements.txt

4. Initialize logger using constructor

src/mlProject/__init__.py

5. Create source code

5.1 Update config.yaml
5.2 Update schema.yaml
5.3 Update params.yaml
5.4 Update the entity
5.5 Update the config
5.6 Update the components
5.7 Update the pipeline
5.8 Update the main.py
5.9 Update the app.py

API Testing (FOR EVERYONE)

Predict = 0

Result: คุณไม่มีความเสี่ยงเป็นโรคเบาหวาน

{
    "age": "50-59",
    "gender": "male",
    "family_diabetes": "no",
    "highbp": "yes",
    "physicallyactive": "one hr or more",
    "bmi": 39,
    "smoking": "no",
    "alcohol": "no",
    "sleep": 8,
    "soundsleep": 6,
    "regularmedicine": "no",
    "junkfood": "occasionally",
    "stress": "sometimes",
    "bplevel": "high",
    "pregancies": 0,
    "pdiabetes": "no",
    "uriationfreq": "not much"
}

Predict = 1

Result: คุณมีความเสี่ยงเป็นโรคเบาหวาน

{
    "age": "60 or older",
    "gender": "male",
    "family_diabetes": "yes",
    "highbp": "yes",
    "physicallyactive": "more than half an hr",
    "bmi": 27,
    "smoking": "no",
    "alcohol": "no",
    "sleep": 6,
    "soundsleep": 5,
    "regularmedicine": "yes",
    "junkfood": "occasionally",
    "stress": "sometimes",
    "bplevel": "high",
    "pregancies": 0,
    "pdiabetes": "no",
    "uriationfreq": "quite often"
}

Model Training (FOR DEVELOPER)

Setup system environment variables

For Unix-like shells (e.g., Bash, macOS Terminal):

export MLFLOW_TRACKING_URI=https://dagshub.com/NutBodyslam053/BD528-software_engineering.mlflow
export MLFLOW_TRACKING_USERNAME=NutBodyslam053
export MLFLOW_TRACKING_PASSWORD=b85bafd69d98861fee89f5bf70dc5f62cf41c2e5

For Windows PowerShell:

$env:MLFLOW_TRACKING_URI = "https://dagshub.com/NutBodyslam053/BD528-software_engineering.mlflow"
$env:MLFLOW_TRACKING_USERNAME = "NutBodyslam053"
$env:MLFLOW_TRACKING_PASSWORD = "b85bafd69d98861fee89f5bf70dc5f62cf41c2e5"

For Windows Command Prompt:

set MLFLOW_TRACKING_URI=https://dagshub.com/NutBodyslam053/BD528-software_engineering.mlflow
set MLFLOW_TRACKING_USERNAME=NutBodyslam053
set MLFLOW_TRACKING_PASSWORD=b85bafd69d98861fee89f5bf70dc5f62cf41c2e5

Run app container

docker run -dp 5000:5000 \
    -e MLFLOW_TRACKING_URI=https://dagshub.com/NutBodyslam053/BD528-software_engineering.mlflow \
    -e MLFLOW_TRACKING_USERNAME=NutBodyslam053 \
    -e MLFLOW_TRACKING_PASSWORD=b85bafd69d98861fee89f5bf70dc5f62cf41c2e5 \
    nutbodyslam053/bd528:v1

MLflow tracking remote: https://dagshub.com/NutBodyslam053/BD528-software_engineering.mlflow

Project Demo

Witchakorn Wanasanwongkot

@witchakornwanasanwongkot9657