Heart Failure Prediction using Neural Networks

This project uses a machine learning approach to predict the survival of patients with heart failure using clinical and demographic features.

Dataset

The dataset is sourced from Kaggle and contains 12 features:

• age: Age of the patient
• anaemia: Presence of anaemia (1 = yes, 0 = no)
• creatinine_phosphokinase: Level of CPK enzyme in the blood (mcg/L)
• diabetes: Presence of diabetes (1 = yes, 0 = no)
• ejection_fraction: Percentage of blood leaving the heart at each contraction
• high_blood_pressure: Presence of hypertension (1 = yes, 0 = no)
• platelets: Platelet count (kiloplatelets/mL)
• serum_creatinine: Level of serum creatinine (mg/dL)
• serum_sodium: Level of serum sodium (mEq/L)
• sex: Gender of the patient (1 = male, 0 = female)
• smoking: Smoking status (1 = yes, 0 = no)
• time: Follow-up period (days)
• death_event: Target variable (1 = death, 0 = survival)

Cardiovascular diseases (CVDs) are the leading cause of death globally, responsible for an estimated 17.9 million deaths annually (31% of all deaths). Heart failure is a major outcome of CVDs, and predicting mortality risk can aid early intervention.

Project Overview

The project implements a neural network classifier using TensorFlow/Keras to predict patient survival based on clinical data. Key steps include:

1. Data Loading and Inspection: Load CSV dataset and inspect structure.
2. Data Preprocessing:
   • One-hot encode categorical variables.
   • Scale numeric features using StandardScaler.
   • Encode target labels for classification.
3. Model Architecture:
   • Input layer matching the number of features.
   • Hidden layer with 12 neurons and ReLU activation.
   • Output layer with 2 neurons (YES/NO) and softmax activation.
4. Training:
   • Loss function: Categorical Crossentropy.
   • Optimizer: Adam.
   • Metrics: Accuracy.
   • Epochs: 100, Batch size: 16.
5. Evaluation:
   • Classification report with precision, recall, f1-score.
   • Accuracy on the test set.

Results

Class	Precision	Recall	F1-score	Support
0	0.74	0.91	0.82	44
1	0.81	0.55	0.65	31
Accuracy	-	-	0.76	75
Macro Avg	0.78	0.73	0.74	75
Weighted Avg	0.77	0.76	0.75	75

Requirements

• Python 3.x
• Pandas
• NumPy
• scikit-learn
• TensorFlow/Keras

Installation

pip install pandas numpy scikit-learn tensorflow