Heart-Failure-Prediction-Using-Machine-Learning-Classification-Algorithms

Dataset

https://www.kaggle.com/datasets/fedesoriano/heart-failure-prediction

Preprocessing Pipeline

1. Numerical features statistics.
2. Missing values detection.
3. Duplicates detection.
4. Target balance detection.

5. Outliers detection and handling.

A) Before handling:

B) After handling:

6. Distribution of numerical and categorical features.

A) Numerical features:

B) Categorical features:

Data Visualization Pipeline

1. Histogram of features between classes.

A) Numerical features:

B) Categorical features:

2. Pairplot of features.

3. Correlation between features.

Feature Scaling Pipeline

1. Quantization of categorical features.

2. Dataset splitting into training, validation and testing.
3. Feature scaling step.

Feature Extraction Pipeline

Backward Sequential Feature Selection

1. Performing backward SFS.
2. GridSearchCV for finding best hyperparameters.
3. Validating models.

Principal Component Analysis

1. Choosing best number of principal components.

2. Data Transformation.
3. GridSearchCV for finding best hyperparameters.
4. Validating models.

Training and Testing Pipeline

1. Training and testing chosen models.
2. Plotting testing results.

3. Visualization of predictions.

A) RF predictions using PCA:

B) RF predictions using SFS:

4. Confusion matrix.