Résultats de recherche pour « Compact Multi-Class Boosted Trees »

Eye is a crucial organ that is responsible for vision and overall quality of life. However, eye diseases such as cataract, glaucoma, and diabetic retinopathy are the leading causes of vision loss and blindness worldwide. Therefore, if people face eye-related problem needs accurate and early-stage diagnosis. The traditional diagnostic process is timeconsuming and not accessible in the rural or remote area. To address this challenge, we develop a hybrid deep learning framework that includes Convolutional Neural Network (CNN), feature extraction, and XGBClassifier. The dataset contains four classes such as cataract, diabetic retinopathy, glaucoma, and normal. In this paper, we first implemented a custom CNN model and evaluated its performance. Next, we used the CNN for deep feature extraction and applied an XGBClassifier on these features, and evaluated its performance. Finally, we developed an ensemble model that combines the first and second steps for improved classification accuracy. Comparative analysis demonstrated that the developed ensemble of CNN and hybrid CNN-XGBClassifier architecture outperformed individual CNN and hybrid CNN-XGBClassifier models. The ensemble model achieved an accuracy of 89.10%, the CNN model achieved 87.56%, and the hybrid CNN-XGBClassifier model achieved 88.74% accuracy score. Additional evaluation using, confusion matrix and classification report confirmed the hybrid model's capability across multiple classes. The findings highlight the potential of the framework to support multi-class eye disease screening.

Early and accurate classification of breast cancer is essential to support clinical diagnostic processes and improve patient outcomes. This study proposes a comprehensive machine learning pipeline based on Gradient Boosted Tree algorithms to classify breast tumors into benign and malignant categories. The proposed framework integrates several preprocessing stages, including outlier handling using the Local Outlier Factor (LOF), feature normalization with StandardScaler, class imbalance handling using SMOTE, and feature selection through ANOVA-based SelectKBest. Five ensemble learning models—XGBoost, LightGBM, CatBoost, HistGradientBoosting, and GradientBoosting—were trained and evaluated using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The experimental results show that all models achieved strong and comparable classification performance. Among them, CatBoost obtained the highest ROC-AUC value of 0.9960, along with an accuracy of 0.9649, precision of 0.9750, recall of 0.9286, and F1-score of 0.9512. Statistical evaluation using the DeLong test indicated that the differences in ROC-AUC among the evaluated models were not statistically significant (p > 0.05), suggesting similar discriminative capabilities across models. To enhance model interpretability, SHAP (SHapley Additive exPlanations) was applied to the CatBoost model as a representative classifier. The results show that features related to nuclear size and shape, such as radius, area, perimeter, and concavity, contributed most significantly to malignant predictions. This study demonstrates that the integration of robust preprocessing techniques, Gradient Boosted Tree models, and explainable machine learning provides an accurate and interpretable approach for breast cancer classification. However, the evaluation was conducted on a single public dataset without external validation, and further studies using independent and real-world datasets are required before clinical deployment.

The early detection and diagnosis of faults in induction motors is vital in today’s industry, since these are the motors used for the largest number of applications in the industrial environment and failure to detect a fault early can lead to significant losses. Bearing faults are the main problems detected in induction motors and several techniques have been developed to detect them. The use of the information contained in the motor vibrations is the main traditional source for its diagnosis, although there are also proposals that use the supply current, or the sound of the motor. Furthermore, these variables can be used in the time domain or in the frequency domain. The purpose of this work is to use explainable artificial intelligence (XAI) to determine which of these variables, and in which domain, contributes most to a correct diagnosis and how much can be gained in diagnosis by using multisensor data fusion. To carry out this comparison in the most objective way possible, a model selection procedure is proposed and boosting techniques are considered that prove to give a very precise diagnosis. The obtained diagnostic rules are then interpreted using SHAP values, a recent interpretation technique for complex classification procedures.

Rapid and reliable identification of respiratory diseases from chest X-ray images remains an important task in computer-aided medical diagnosis. This study proposes a lightweight, ensemble-based boosting framework for multi-class classification of chest X-ray images into COVID-19, pneumonia, and normal categories. The proposed method integrates three efficient convolutional neural network architectures, namely DenseNet121, MobileNetV2, and NASNetMobile, and combines their outputs through a boosted ensemble mechanism to improve predictive performance while preserving computational efficiency. The dataset was split into training, validation, and test sets, and the images were resized, normalised, and augmented prior to model training. Experimental results showed that the proposed boosted ensemble model achieved a classification accuracy of 97.40% , outperforming the individual base models DenseNet121 ( 91.60% ), MobileNetV2 ( 94.40% ), and NASNetMobile ( 93.46% ). These findings indicate that boosted lightweight ensemble learning can provide an effective and practical solution for automated chest X-ray image classification in medical imaging applications.