jordanflood / ml-prediction-of-cvd

ML Prediction of CVD

By Jordan Flood

Data Analysis and Predictive Modeling of Cardiovascular Disease (CVD)

Overview

The project "Data Analysis of Cardiovascular Disease (CVD)" aims to analyze the factors contributing to cardiovascular disease and build a machine-learning model to predict the likelihood of an individual developing CVD. The analysis was performed using a dataset containing over 70,000 individual records from the Kaggle platform. The findings provide insights into various lifestyle, genetic, and environmental factors that play a role in cardiovascular health.

Dataset

• Source: Cardiovascular Disease Dataset (Kaggle)
• Size: 70,000 individuals
• Objective: To discover the factors influencing CVD and build a predictive model

Process Flow

1. Discovery & Objective Definition

   • Objective: Identify key factors contributing to CVD and build a predictive model

2. Data Cleaning

   • Handling missing values and outliers
   • Normalization and scaling of numerical data

3. Exploratory Data Analysis

   • Identify correlations between variables and CVD
   • Visualize the relationships using various graphs

4. Model Building & Testing

   • Build machine learning models to predict CVD
   • Optimize and evaluate the model performance

5. Results Interpretation & Reporting

   • Interpret key findings and draw actionable recommendations
   • Visualize insights and communicate results effectively

Key Findings

1. Age:

   • Increased CVD cases are noticeable between ages 50-55.
   • Aging causes heart muscles to thicken and arteries to stiffen, increasing blood pressure.

2. Body Mass Index (BMI):

   • Higher BMI is associated with a higher risk of CVD.
   • Extra strain on the heart due to high BMI leads to insulin resistance and type 2 diabetes.

3. Blood Pressure:

   • Elevated blood pressure (hypertension) is a major risk factor for CVD.
   • Increased workload due to hypertension can damage artery walls and cause heart attacks.

4. Lifestyle:

   • Physical inactivity, smoking, and heavy alcohol consumption contribute to increased CVD risk.

Model Development & Evaluation

1. Model Building:

   • The dataset was used to train a machine-learning model to predict CVD.
   • Features used included age, BMI, blood pressure, cholesterol, and lifestyle factors.

2. Model Performance:

   • The predictive model achieved an accuracy of approximately 72%.

3. Feature Importance:

   • Age, BMI, and blood pressure were the most significant features.

Recommendations

1. Enhanced Public Awareness:

   • Promote lifestyle changes like a balanced diet and regular physical activity.
   • Early screening and intervention should be encouraged.

2. Integration into Clinical Practice:

   • Incorporate predictive models in routine clinical assessments.

3. Lifestyle Interventions:

   • Expand programs to reduce risk factors through accessible lifestyle changes.

4. Expand Sample Size:

   • Use a larger and more diverse sample size for better generalizability.

5. Incorporate More Variables:

   • Add more detailed variables, such as specific cholesterol levels (LDL, HDL).

Repository Structure

/
|-- notebooks/
|   |-- eda-CVD-final.ipynb
|-- data/
|   |-- Data Analysis of Cardiovascular Disease.pdf
|-- README.md