Google Advanced Data Analytics — Capstone Project

The culminating project of the Google Advanced Data Analytics Professional Certificate, featuring three end-to-end ML scenarios using the PACE (Plan, Analyze, Construct, Execute) framework.

Scenarios

#	Scenario	Business Problem	ML Approach
1	Automatidata	NYC TLC fare/tip prediction	Linear Regression, Random Forest, XGBoost regression
2	TikTok	Claims vs. Opinions classification	Logistic Regression, Random Forest, XGBoost classifier
3	Waze	User churn prediction	Random Forest (grid search), XGBoost classifier

Project Structure

google-ada-capstone-project/
├── src/
│   ├── pipeline.py            # Reusable MLPipeline class (preprocess → train → evaluate)
│   └── config.py              # Scenario configurations
├── notebooks/
│   ├── automatidata/
│   │   └── capstone_automatidata.ipynb
│   ├── tiktok/
│   │   └── capstone_tiktok.ipynb
│   └── waze/
│       └── capstone_waze.ipynb
├── data/                      # CSV datasets
├── results/                   # Saved metrics and visualizations
├── requirements.txt
└── README.md

PACE Framework

Each notebook follows the structured approach taught in the certificate:

1. Plan — Define the business problem, stakeholders, and success metrics
2. Analyze — EDA: distributions, correlations, missing data, outliers
3. Construct — Feature engineering, model training, hyperparameter tuning
4. Execute — Evaluation, interpretation, and business recommendations

Reusable Pipeline

from src.pipeline import MLPipeline

pipe = MLPipeline(df, target="churn", features=feature_list)
pipe.preprocess(scale=True)
pipe.train("xgboost")
pipe.evaluate()
pipe.feature_importance()
pipe.cross_validate(cv=5)

Getting Started

git clone https://github.com/asenabeshiktepeli/google-ada-capstone-project.git
cd google-ada-capstone-project
pip install -r requirements.txt

# Run any capstone notebook
jupyter notebook notebooks/automatidata/capstone_automatidata.ipynb

Key Results

• Automatidata: XGBoost regression achieves strong R² for sales prediction; TV is the #1 driver
• TikTok: XGBoost classifier with high AUC; online boarding and entertainment top predictors
• Waze: Random Forest with grid search yields best F1 for churn prediction; efficiency and points key features

Technologies

• Python 3.10+
• scikit-learn, XGBoost, scipy
• pandas, NumPy, matplotlib, seaborn