Predictive Modeling for Fetal Health: A Study of PCA Dimensionality Reduction

Predictive modeling in fetal health is essential for understanding the condition and development of a fetus. This project explores the application of machine learning algorithms to classify fetal health into three categories: Normal, Suspect, and Pathological. The study focuses on the use of Principal Component Analysis (PCA) for dimensionality reduction to improve computational efficiency and model performance. Four classification algorithms are implemented: Naive Bayes, Logistic Regression, Multi-Layer Perceptron (MLP), and Random Forest Classifier. The dataset contains multiple features derived from cardiotocography recordings, providing detailed information on fetal heart rate and uterine activity. Data preprocessing steps include handling missing values, feature scaling, and applying PCA to reduce redundant information. The models are evaluated using accuracy, precision, recall, and F1-score to determine their effectiveness in fetal health classification. Comparative analysis highlights how PCA influences prediction performance and identifies the most suitable algorithm for the dataset. The system is deployed as a web application using Flask, with modules for Home, Register, Login, Classification, and Logout, providing a simple interface for interacting with the predictive models. The study demonstrates that dimensionality reduction can enhance model efficiency without significantly affecting accuracy, offering a structured approach for fetal health assessment using machine learning.

Keywords: Fetal health, Predictive modeling, PCA, Dimensionality reduction, Naive Bayes, Logistic regression, Random Forest, MLP, Classification, Machine learning.