1. Abstract

Hospital mortality prediction is an important task in healthcare analytics, especially for patients admitted to Intensive Care Units (ICUs). Predicting whether a patient is likely to survive or not based on clinical parameters can help doctors take preventive measures and improve treatment planning. However, traditional ICU prediction systems often lack accuracy and fail to capture complex patterns in patient health data.

This project focuses on predicting hospital mortality using machine learning techniques and Automated Machine Learning (AutoML) with the PyCaret library. The dataset contains various patient health attributes and medical parameters collected from ICU records.

Data preprocessing techniques such as handling missing values, feature selection, and normalization are applied to prepare the dataset. Machine learning models are first implemented manually to understand model behaviour. Then, PyCaret AutoML is used to automatically compare multiple models, optimize hyperparameters, and select the best performing model.

The final model predicts the probability of patient survival or mortality based on input health parameters. This project demonstrates how machine learning and AutoML can assist healthcare professionals in improving clinical decision-making and patient outcome prediction.

2. Objectives

The main objectives of this project are:

1. To understand the problem of hospital mortality prediction in ICU patients.
2. To analyse medical datasets and identify important health parameters.
3. To perform data preprocessing and feature engineering.
4. To build machine learning models for mortality prediction.
5. To study the concept of Automated Machine Learning (AutoML).
6. To implement the PyCaret library for automated model selection.
7. To evaluate model performance using classification metrics.
8. To develop an efficient system for predicting patient mortality risk.

3. Existing System

Traditional methods used for hospital mortality prediction include:

• Clinical judgement based on doctor experience

• Statistical models used in healthcare analytics

• Basic machine learning algorithms applied manually

Limitations of Existing Systems

1. Low prediction accuracy in complex medical conditions.
2. Difficult to capture hidden patterns in large healthcare datasets.
3. Time-consuming model development and tuning process.
4. Manual model selection may miss better performing algorithms.
5. Limited automation in model optimization and evaluation.

These limitations highlight the need for automated and intelligent prediction systems using AutoML.

4. Proposed System

The proposed system predicts hospital mortality using machine learning and AutoML techniques with the PyCaret library.

In this system:

• ICU patient dataset containing medical attributes is collected.

• Data preprocessing and feature engineering are performed.

• Multiple machine learning models are trained for prediction.

• PyCaret AutoML automatically compares various models.

• The best performing model is selected and optimized.

• The system predicts whether a patient is likely to survive or not based on medical parameters.

This system improves prediction accuracy and reduces the time required for model development.

 5. Implementation Procedure

The implementation of this project includes the following steps:

Step 1: Data Collection

The hospital mortality dataset is obtained from publicly available medical datasets containing ICU patient information.

Step 2: Data Preprocessing

The dataset is processed by:

• Handling missing values

• Removing irrelevant features

• Encoding categorical variables

• Normalizing the dataset

• Splitting the data into training and testing sets

Step 3: Exploratory Data Analysis (EDA)

• Visualizing relationships between patient parameters

• Understanding data distribution

• Identifying important features affecting mortality

Step 4: Model Development

Machine learning algorithms such as:

• Logistic Regression

• Decision Tree

• Random Forest

• Support Vector Machine

are implemented to predict patient mortality.

Step 5: AutoML Implementation Using PyCaret

• PyCaret library is used to automate model comparison.

• Multiple models are tested automatically.

• Hyperparameters are tuned automatically.

• The best performing model is selected.

Step 6: Model Evaluation

Model performance is evaluated using metrics such as:

• Accuracy

• Precision

• Recall

• F1 Score

• Confusion Matrix

Step 7: Prediction System

The trained model is used to predict mortality risk by inputting patient health parameters.

6. Software Requirements

The software tools used in this project include:

• Python – Programming language

• Jupyter Notebook / Google Colab – Development environment

• PyCaret – Automated Machine Learning library

• NumPy – Numerical computation

• Pandas – Data manipulation

• Matplotlib / Seaborn – Data visualization

• Scikit-learn – Machine learning utilities

7. Hardware Requirements

Minimum Hardware Requirements:

• Processor: Intel i5 or higher

• RAM: 8 GB or higher

• Storage: 256 GB or higher

• Laptop or Desktop Computer

• Internet connection for dataset download

 8. Advantages of the Project

1. Improves accuracy of hospital mortality prediction.
2. Helps doctors make better clinical decisions.
3. Automates model development using AutoML.
4. Reduces time required for model selection and tuning.
5. Can analyse large healthcare datasets efficiently.
6. Useful for early risk detection in ICU patients.
7. Demonstrates practical application of machine learning in healthcare.