Osteoporosis Risk Prediction

Osteoporosis is a bone disease that leads to an increased risk of fractures due to decreased bone density. It is particularly prevalent among older adults and postmenopausal women.Early prediction of osteoporosis risk can help in preventive care and reduce healthcare costs.

This project aims to predict the risk of osteoporosis in patients based on their health data using machine learning. The project follows the entire machine learning lifecycle, from data exploration and preprocessing to model training, evaluation, and deployment.

Project Overview

• Objective: Predict the risk of osteoporosis in patients using health data.
• Target Variable: Binary classification (1 = At Risk, 0 = Not At Risk).
• Model: XGBoost (Extreme Gradient Boosting).
• Evaluation Metrics: Accuracy, Precision, Recall, F1-Score, ROC-AUC.

---

Dataset The dataset used in this project is osteoporosis.csv. It contains 1958 rows (patients) and 16 columns (features). Below is a description of the features:

Feature	Description
Id	Unique identifier for each patient
Age	Age of the patient (in years)
Gender	Gender of the patient (Male/Female)
Hormonal Changes	Hormonal status (e.g., Normal, Postmenopausal)
Family History	Family history of osteoporosis (Yes/No)
Race/Ethnicity	Race or ethnicity of the patient
Body Weight	Body weight category (e.g., Underweight, Normal)
Calcium Intake	Daily calcium intake (Low/Adequate)
Vitamin D Intake	Vitamin D intake level (Insufficient/Sufficient)
Physical Activity	Level of physical activity (Sedentary/Active)
Smoking Status	Smoking status of the patient (Smoker/Non-smoker)
Alcohol Consumption	Alcohol consumption level (None/Moderate/Heavy)
Medical Conditions	Existing medical conditions (e.g., Rheumatoid Arthritis)
Medications	Medications taken (e.g., Corticosteroids)
Prior Fractures	History of prior fractures (Yes/No)
Osteoporosis	Target variable (1 = At Risk, 0 = Not At Risk)

Installation

1. Clone the repository: git clone https://github.com/Hanah29/machin-learning.git cd machin-learning

2. Install the required dependencies: pip install -r requirements.txt

Usage

1. Data Exploration and Preprocessing:

   • Run the Jupyter Notebook data_exploration.ipynb to explore the dataset and perform preprocessing steps.

2. Model Training and Evaluation:

   • Run the Python script train_model.py to train the XGBoost model and evaluate its performance. python train_model.py

3. Model Deployment (Optional):

   • Deploy the trained model as an API using FastAPI: uvicorn app:app --reload

   • Test the API using Postman or cURL: curl -X POST "http://127.0.0.1:8000/predict" -H "Content-Type: application/json" -d '{"Age": 65, "BMI": 25, ...}'

Results

• Model Performance:

  • Accuracy: 0.8622
  • Precision: 0.9277
  • Recall: 0.7857
  • F1-Score: 0.8508
  • ROC-AUC: 0.8910

• Comparison Against Baseline:

  • Baseline Accuracy: 0.5000 (Dummy Classifier)
  • XGBoost Accuracy: 0.8622

Deployment The trained model can be deployed as an API using FastAPI. Follow these steps to deploy and test the API:

1. Run the FastAPI server: uvicorn app:app --reload

2. Test the API using Postman or cURL: curl -X POST "http://127.0.0.1:8000/predict" -H "Content-Type: application/json" -d '{"Age": 65, "BMI": 25, ...}'

Author Hana Ayalew Mhiret id -DBU1401373

License This project is licensed under the MIT License.