🫀 Cardiovascular Disease Project

Business Understanding

The company

Cardio Catch Disease is a healthcare company specialized in diagnosing cardiovascular disease in early stages. Its bussiness model is offering this early diagnosis of cardiovascular disease for a certain price.

The problem

Currently, the precision of the diagnosis ranges between 55% and 65%, reflecting its complexity and the limited availability of medical doctors. The total cost per diagnosis, encompassing both equipment and analysts' compensation, is approximately $1,000.

For clients, the pricing model for the diagnosis costs $500 for every 5% increase in precision beyond the 50% baseline.

Objective

Create a model that improves the precision of the diagnosis in a stable and efficient way for all clients.

What to deliver

• The precision of the model
• What is the revenue of the new model?
• How reliable are the results of the new tool?

Solution Planning

Dataset from Kaggle

CRISPDM Process

- Business understanding
- Data understanding
- Data preparation
- Data analysis
- Feature engineering
- Modeling
- Model development
- Model evaluation
- Model deployment

Project Structure

- Imports
- Data Description
- Preprocessing and Data Exploration
- Exploratory Data Analysis
- Feature Engineering
- Machine Learning and Evaluation
- Business Results

Steps

• Imports

Importing libraries and packages (pandas, numpy, seaborn, matplotlib, boruta, sklearn, lightgbm), configuring settings, loading dataset.

• Data Description

Shape, dimensions, info, describe, columns, types and head of the dataset.

• Preprocessing and Data Exploration

Transforming age from days to years, checking data balance, describing numerical data, corrections on the dataset.

• Exploratory Data Analysis

Answering questions with graphs and charts, like:

At what age people with cardiovascular disease outnumber people without cardiovascular disease? Are older people more at risk?

We can see that older people are more at risk and that at 55 years old, people with cardiovascular disease outnumber the people without.

Occurences of features in the dataset:

The higher the cholesterol, the greater the risk?

The higher the cholesterol, the greater the risk of having a cardiovascular disease.

• Feature Engineering

Dropping columns, creation of features like body-mass-index (BMI), correlation heatmap, train-test split, scaling with MinMaxScaler and Robust Scaler, features selection with ExtraTreesClassifier and Boruta.

• Machine Learning and Evaluation

  • Machine Learning models: Logistic Regression, Extra Trees Classifier, Random Forest Classifier, LightGBM Classifier.

  • Evaluation metrics: Accuracy, Precision, Recall, F1 Score, Confusion Matrix.

  • Hyperparameters tuning: RandomizedSearchCV on Random Forest Classifier and LightGBM Classifier.

Business Results

Reminder:

Currently, the precision of the diagnosis ranges between 55% and 65%, reflecting its complexity and the limited availability of medical doctors. The total cost per diagnosis, encompassing both equipment and analysts' compensation, is approximately $1,000.

For clients, the pricing model for the diagnosis costs $500 for every 5% increase in precision beyond the 50% baseline.

The deliverables:

• The precision of the model now is at 75%

• The revenue of the new model is $173,665,751.75

• Considering the previous precision of 65%, the revenue was around $105,000,000, which results in an improvement of $68,665,751.75 (or an increase of 65%) using the new model.

• How reliable are the results of the new tool? The other metrics for the model are Accuracy: 72.3%, Recall: 65.8%, F1 Score: 70%.

---