Pregnancy Prediction Analysis

This project analyzes purchasing behavior patterns to predict pregnancy status using logistic regression. The model achieves 84.5% accuracy in predicting pregnancy based on various shopping behaviors and lifestyle changes.

Installation

1. Clone the repository:

git clone https://github.com/jfm56/regression_analysis.git
cd regression_analysis

2. Install required dependencies:

pip install -r requirements.txt

Required dependencies:

• pandas==2.1.0
• numpy==1.24.3
• scikit-learn==1.3.0
• openpyxl==3.1.2
• matplotlib==3.7.1
• seaborn==0.12.2

Development dependencies:

• pytest==7.4.0
• pylint==2.17.5
• pytest-cov==4.1.0
• black==23.7.0

Usage

Data Format

The analysis expects an Excel file (target.xlsx) with the following columns:

• Implied Gender
• Home/Apt/PO Box
• Pregnancy Test
• Birth Control
• Feminine Hygiene
• Folic Acid
• Prenatal Vitamins
• Prenatal Yoga
• Body Pillow
• Ginger Ale
• Sea Bands
• Stopped buying ciggies
• Cigarettes
• Smoking Cessation
• Stopped buying wine
• Wine
• Maternity Clothes
• Pregnant (target variable, 0 or 1)

An example file target.xlsx is included in the repository.

Running the Analysis

1. Place your data file (Excel format) in the project directory
2. Run the analysis:

python -m regression_analysis.regression

The script will generate three visualization files:

• correlation_heatmap.png: Shows correlations between features
• scatter_plots.png: Displays relationships between key features and pregnancy
• feature_importance.png: Shows the importance of each feature

Analysis Results

Model Performance

• Accuracy: 84.50%
• Precision (Pregnant): 92%
• Recall (Pregnant): 74%
• F1-Score (Pregnant): 82%

Key Findings

Strongest Positive Indicators of Pregnancy:

1. Folic Acid (2.94)
2. Prenatal Vitamins (2.22)
3. Pregnancy Test (1.96)
4. Maternity Clothes (1.72)
5. Ginger Ale (1.41)

Strongest Negative Indicators:

1. Birth Control (-2.03)
2. Feminine Hygiene (-1.73)
3. Wine (-1.29)
4. Cigarettes (-1.25)

Visualization Explanation

Scatter Plots

The scatter plots show the relationships between six key features and pregnancy status:

• X-axis: Feature value
• Y-axis: Pregnancy status (0 = Not Pregnant, 1 = Pregnant)
• Red trend lines indicate the direction and strength of relationships
• Upward trends suggest positive correlation with pregnancy
• Downward trends suggest negative correlation with pregnancy

Key observations:

• Folic Acid and Prenatal Vitamins show strong positive correlations
• Birth Control and Wine show strong negative correlations
• The spread of points indicates the reliability of each relationship

Feature Importance

The model identifies purchasing patterns that are most predictive of pregnancy:

• Health supplements (Folic Acid, Prenatal Vitamins) are the strongest positive indicators
• Contraceptives and lifestyle products (Birth Control, Wine, Cigarettes) are strong negative indicators
• Changes in purchasing behavior (stopping wine/cigarettes) are moderately strong indicators

Development Setup

Local Development

1. Install development dependencies:

pip install -r requirements.txt

2. Run tests:

pytest

3. Check code quality:

pylint regression_analysis/regression.py tests/*.py

4. Check test coverage:

pytest --cov=regression_analysis --cov-report=html

The test suite includes:

• Model accuracy testing with realistic data
• Feature importance validation
• Error handling tests
• 90% code coverage

Docker Development

Using Pre-built Image

docker pull jmullen029/regression_analysis:latest
docker run jmullen029/regression_analysis:latest

Local Development with Docker

1. Build and run using Docker:

docker-compose up --build

2. Run tests in Docker:

docker-compose run regression pytest

The Docker image is automatically built and published to Docker Hub on every push to main branch.

Notes

• The model uses logistic regression for binary classification (Pregnant/Not Pregnant)
• Features are encoded using Label Encoding for categorical variables
• The dataset is split 80/20 for training and testing
• Results include both positive and negative predictors for comprehensive analysis
• Includes comprehensive test suite with pytest
• Docker support for consistent development environment
• Code quality maintained with pylint
• Test coverage tracked with pytest-cov
• Automated dependency updates with Dependabot
  • Weekly checks for Python packages
  • Weekly checks for GitHub Actions
  • Weekly checks for Docker base images
  • Auto-merges patch updates