Evaluation of Machine Learning and Ensemble Learning Models for Classification Using Delivery Data

Purpose: This study aims to evaluate the performance of various machine learning and ensemble learning models in classifying delivery times using Amazon delivery data. Fast deliveries' role in providing a competitive advantage and boosting customer loyalty highlights the importance of this study. Methodology: The research employs a dataset of 43,739 delivery records with 15 features. Data preprocessing steps include handling missing values, encoding categorical variables, calculating geospatial distances, and normalizing data. Advanced machine learning techniques (e.g., KNN, SVM, Logistic Regression) and ensemble methods (e.g., ExtraTrees, AdaBoost) were systematically compared based on accuracy, precision, recall, and F-score. Findings: Ensemble learning models, particularly those using SVM, NB, and LDA as base models and ET as the meta model, achieved the highest accuracy (99.89%) and F-score (99.89%). These results underscore the potential of such models to optimize logistics operations, reduce delays, and enhance customer satisfaction. Originality: This study demonstrates the effectiveness of machine and ensemble learning methods on complex logistics data, contributing to optimizing logistics efficiency and enhancing customer satisfaction. Additionally, the application of ensemble learning methods on complex and large-scale logistics data structures is unique in terms of its contribution to the literature. The proposed framework offers a scalable solution for real-time predictive modeling and logistics optimization.