Hybrid Model for Nail Health Classification Using Decision Tree, Gradient Boosting and KNN.

This research paper introduces a Hybrid Deep Learning Method for Nail Health Classification that can precisely differentiate between healthy and unhealthy nails. The system uses a Convolutional Neural Network (CNN) for feature extraction, taking advantage of its capability to detect detailed patterns like edges, textures, and shapes necessary for medical image analysis. After feature extraction, three different classifiers—Decision Tree, Gradient Boosting, and K-Nearest Neighbors (KNN)—are used to carry out the classification process. These classifiers were chosen judiciously based on their complementary strengths: Decision Tree for interpretability and low computational cost, Gradient Boosting for high accuracy and robustness against overfitting, and KNN for its ability to capture local patterns in the data. The dataset includes images of healthy and diseased nails, divided into training and validation sets to provide strong model evaluation. For consistency, all images were resized to 150x150 pixels, and data augmentation methods were used to improve model generalization. The CNN model was built with three convolutional layers and max-pooling layers, ending with a fully connected layer for feature extraction. These are the high-level features which were then passed on as inputs to the three classifiers, and all three classifiers were trained and tested upon the same dataset. The models scored individually 87.57% for Decision Tree, 87.57% for Gradient Boosting, and 88.65% for KNN. In order to improve overall classification performance, the results of all three classifiers were combined through majority voting to create a hybrid model. This ensemble method takes advantage of the diversity of the classifiers by efficiently balancing their strengths while reducing their individual weaknesses. The hybrid model performed better than each individual classifier, proving the efficacy of this combined strategy. A user-friendly interface was created to present the input nail image with its classification output, allowing for real-time assessment of nail health. The interface not only promotes ease of use but also the detection of disease at an early stage, thus preventing unnecessary doctor consultations. The models, that is, the CNN and the three classifiers, were stored in deployable forms (.h5 and.pkl files) for ease of use and scaling up for subsequent studies or medical purposes. This work makes an addition to the area of medical image analysis by proving the efficiency of integrating CNN-based feature extraction with various classifiers within a combined framework. The system presented is a sound, precise, and efficient automated nail health screening solution, with its potential impact seen in public health, particularly in resource-limited environments. The findings highlight the feasibility of employing artificial intelligence in health diagnostics, making way for continued innovation in automatic disease detection and medical decision-making systems.