Gaussian Naive Bayes-based Heart Disease Prediction System: Practical Application of Machine Learning in Medical Diagnosis

This article introduces a heart disease prediction system built using the Gaussian Naive Bayes algorithm. Based on medical data from 918 patients, it achieves an accuracy rate of 85.3%, providing an efficient machine learning solution for early heart disease screening and demonstrating the practical value of machine learning in the field of medical diagnosis.

Project Background and Significance

Heart disease is one of the leading causes of death worldwide. According to the World Health Organization, cardiovascular diseases cause approximately 17.9 million deaths each year, accounting for 32% of global deaths. Early detection and intervention are key to reducing heart disease mortality. However, traditional diagnostic methods often rely on doctors' experience and expensive examination equipment, which is particularly challenging in areas with limited medical resources.

The rise of machine learning technology has brought new possibilities to medical diagnosis. By analyzing large amounts of patient data, machine learning models can identify potential disease risk patterns and assist doctors in making faster and more accurate diagnostic decisions. This project is based on this concept and builds a lightweight but efficient heart disease prediction system.

Dataset Overview and Feature Engineering

The dataset used in this project contains medical records of 918 patients, covering 12 key medical indicators. These indicators include basic patient information (age, gender), symptom manifestations (chest pain type, exercise-induced angina), physiological indicators (resting blood pressure, cholesterol level, maximum heart rate), and ECG-related data (resting ECG results, ST segment slope).

Data preprocessing is a key step for model success. The project team first cleaned the data, handling missing values and outliers. Then, categorical variables were converted to numerical format through label encoding to make them processable by machine learning algorithms. Notably, the project also performed feature engineering, creating a composite indicator called "risk score" that combines cholesterol levels and resting blood pressure to better capture the patient's comprehensive cardiovascular risk.

Selection and Principle of Gaussian Naive Bayes Algorithm

Among numerous machine learning algorithms, this project selected Gaussian Naive Bayes as the core prediction model. This choice is based on the following considerations:

First, the Naive Bayes algorithm is computationally efficient, especially suitable for small to medium-sized datasets. For medical application scenarios, fast response capability is crucial, especially in scenarios where a large number of patients need to be screened in real time.

Second, the algorithm is based on probability theory and can not only provide prediction results but also confidence assessments. This probabilistic output is particularly important for medical decision support, as doctors can judge whether further examinations are needed based on the confidence level.

Gaussian Naive Bayes assumes that features follow a Gaussian distribution (normal distribution). It calculates the conditional probability of each feature under different categories, and combines Bayes' theorem to derive the posterior probability, thereby achieving classification prediction. Although the "naive" independence assumption often does not hold in reality, the algorithm can still achieve satisfactory results in many practical applications.

Model Training and Evaluation Results

The project divided the dataset into training and test sets in an 80/20 ratio. During training, the model learned the statistical correlations between various features and heart disease. The trained model performed well on the test set, achieving the following evaluation metrics:

• Accuracy: 85.3% — the proportion of overall correct predictions
• Precision: 85.1% — the proportion of samples predicted as heart disease that are actually ill
• Recall: 87.8% — the proportion of actually ill samples correctly identified
• F1 Score: 86.4% — the harmonic mean of precision and recall

From the confusion matrix, the model performed well in identifying real patients (recall rate of 87.8%), which means the system can effectively capture most heart disease cases and reduce the risk of missed diagnosis. Among 86 actual patients, the system correctly identified 74 cases, with only 12 misjudged as healthy.

Technical Implementation and Deployment

The project's technology stack is centered on Python, combined with mainstream tools for data science and web development. The main dependencies include:

• Pandas: Used for data processing and cleaning
• NumPy: Provides numerical computing support
• Scikit-learn: Implements machine learning algorithms and model evaluation
• Seaborn: Used for data visualization
• Streamlit: Builds interactive web application interfaces

The project provides a complete code implementation, including the entire process of data loading, preprocessing, model training, prediction, and evaluation. Through the Streamlit framework, developers can quickly build a user-friendly web interface, making it easy for non-technical personnel to use the prediction tool.

Application Prospects and Limitations

The heart disease prediction system has broad application potential. In primary medical institutions, it can serve as a preliminary screening tool to help doctors quickly identify high-risk patients and optimize the allocation of medical resources. In the field of health management, the system can be integrated into personal health monitoring applications to provide users with personalized health risk assessments.

However, it should be clear that machine learning prediction systems should be used as auxiliary tools rather than diagnostic bases. The model's prediction results need to be combined with professional doctors' clinical judgments and verified with necessary medical examinations. In addition, the model's performance is limited by the representativeness and quality of the training data, and its generalization ability in different populations or medical environments still needs further verification.

Summary and Outlook

The Gaussian Naive Bayes-based heart disease prediction system demonstrates the practical value of machine learning in the field of medical diagnosis. Using simple algorithms and public datasets, the project achieved a prediction accuracy of over 85%, proving that even basic machine learning technologies can play an important role in specific scenarios.

In the future, the project can be expanded in multiple directions: introducing more features (such as lifestyle, family medical history), trying more complex models (such as ensemble learning, deep learning), and conducting verification and optimization in actual clinical environments. With the accumulation of medical data and the progress of algorithms, machine learning will surely play an increasingly important role in precision medicine and disease prevention.