Title : Enhancing Model's Ability in Predicting Heart Disease with NCL Algorithm and GridSearchCV
About : Predicting Heart Disease of a Patient or An Individual Using Logistic Regression
Project Description : This portfolio showcases a project focused on enhancing the ability to predict heart disease using machine learning techniques. Heart disease is a global health issue with a high mortality rate, and early detection plays a vital role in reducing the number of deaths.
The project begins with a comprehensive background analysis of the challenges in predicting heart disease accurately. It highlights the lack of awareness among patients regarding early symptoms and the unfortunate cases where many deaths occur due to heart attacks. To address this, the project utilizes machine learning algorithms as a tool for early detection and prediction of heart disease.
The portfolio emphasizes the significance of data mining and machine learning in processing complex healthcare data. Specifically, it explores the application of the Logistic Regression algorithm, a popular classification algorithm, to build a prediction model. The dataset used for training and evaluation is obtained from Kaggle, sourced from the Centers for Disease Control and Prevention (CDC).
To enhance the model's performance, several stages are undertaken. First, data pre-processing techniques are implemented to ensure the quality and reliability of the dataset. Next, the model undergoes optimization through the utilization of the Neighborhood Cleaning Rule (NCL) algorithm. This approach addresses the challenge of imbalanced class distribution commonly found in medical datasets. Additionally, hyperparameter tuning using the GridSearchCV method further improves the adaptability and effectiveness of the Logistic Regression model.
Evaluation of the model's predictive ability is performed using various metrics, including accuracy, confusion matrix, area under the curve (AUC), and recall. The results demonstrate a substantial enhancement in the model's performance compared to the initial state. The recall score, indicating the model's ability to correctly identify positive cases of heart disease, increases from 0.10 to an impressive 0.92, while the AUC improves from 0.54 to 0.945.
Finally, the portfolio highlights the practical implementation of the developed model on a website. This web-based platform allows users to input new data for prediction, providing instant results regarding the presence or absence of heart disease in an individual. By leveraging the power of machine learning, this portfolio project aims to contribute to early detection and prevention efforts, ultimately improving patient outcomes in the fight against heart disease.