This study evaluates the performance of Decision Tree, AdaBoost, and Random Forest models on the case study dataset to address challenges in predictive modeling. The research highlights the limitations of single decision trees, including high variance and overfitting, and explores ensemble methods as solutions. The objective is to compare these algorithms through accuracy, precision, recall, and F1-score metrics. A rigorous methodology involving data preprocessing, hyperparameter tuning, and 5-fold cross-validation was employed to ensure robust evaluation. Results indicate that Random Forest outperformed with an accuracy of 80% and an F1-score of 71.9%, leveraging feature interdependencies effectively. AdaBoost achieved 75% accuracy, balanced precision-recall but was sensitive to noise, while Decision Trees provided 76% accuracy, interpretability but struggled with generalization. Future research should explore advanced ensemble techniques like XGBoost and feature engineering strategies to enhance model performance. These findings underscore the importance of selecting appropriate algorithms tailored to dataset characteristics for real-world applications.

This study evaluates the performance of Decision Tree, AdaBoost, and Random Forest models on the case study dataset to address challenges in predictive modeling. The research highlights the limitations of single decision trees, including high variance and overfitting, and explores ensemble methods as solutions. The objective is to compare these algorithms through accuracy, precision, recall, and F1-score metrics. A rigorous methodology involving data preprocessing, hyperparameter tuning, and 5-fold cross-validation was employed to ensure robust evaluation. Results indicate that Random Forest outperformed with an accuracy of 80% and an F1-score of 71.9%, leveraging feature interdependencies effectively. AdaBoost achieved 75% accuracy, balanced precision-recall but was sensitive to noise, while Decision Trees provided 76% accuracy, interpretability but struggled with generalization. Future research should explore advanced ensemble techniques like XGBoost and feature engineering strategies to enhance model performance. These findings underscore the importance of selecting appropriate algorithms tailored to dataset characteristics for real-world applications.

Urban water infrastructure faces increasing deterioration, necessitating accurate, cost-effective condition assessment. Traditional inspection techniques are intrusive and inefficient, creating demand for scalable machine learning (ML) solutions. This study develops a hybrid ML meta-model to predict underground pipe conditions using a comprehensive dataset of 11,544 records. The objective is to enhance multi-class classification performance while preserving interpretability. A stacked hybrid architecture was employed, integrating Random Forest, LightGBM, and CatBoost models. Following data preprocessing, feature engineering, and correlation analysis, the neural network-based stacking meta-model achieves 96.67% accuracy, surpassing individual base learners while delivering enhanced robustness through model diversity, improved probability calibration, and consistent performance on challenging intermediate condition classes, which are essential for condition prioritization. Age emerged as the most influential feature, followed by length, material type, and diameter. ROC-AUC scores ranged from 0.894 to 0.998 across all models and classes, confirming high discriminative capability. This work demonstrates hybrid architectures for infrastructure diagnostics.

This study evaluates machine learning methods, focusing on neural networks, to address the XOR classification problem and taxi trip duration prediction in New York City. Addressing the need for robust predictive models, the research developed a custom neural network library with features like forward and backward propagation, Sigmoid and Tanh activation functions, and binary cross-entropy and mean squared error loss functions. The XOR model achieved a significant reduction in loss from 0.7389 to 0.0154 over 10,000 epochs, with predictions closely approximating expected outputs. For the taxi trip duration task, three architectures—small (RMSLE: 0.474), medium (RMSLE: 0.455), and large (RMSLE: 0.462)—were evaluated, with the medium model achieving the best balance of performance and efficiency. These results highlight the importance of feature engineering and architecture selection in achieving accurate predictions. Future research should explore advanced techniques like Neural Architecture Search (NAS) and hybrid models to enhance scalability across fields such as transportation, healthcare, environmental impact, and energy optimization.