Classical and Modern Machine Learning Classification Algorithms: Analysis and Comparison

Abstract

With an emphasis on both classic and deep learning models, the research work compares the performance effectiveness of machine learning algorithms in picture classification. Ensemble approaches, such as XGBoost and random forest, were assessed, along with algorithms including support vector machines (SVM), k-nearest neighbors (k-NN), decision trees, random forests, and convolutional neural networks (CNNs). The study utilizes publicly accessible datasets, such as CIFAR-10, and evaluates the results using measures including accuracy, precision, recall, and F1-score. In terms of accuracy and F1-scores, deep learning models—in particular, CNNs and ResNet—performed better than conventional models, proving their effectiveness in managing huge and intricate datasets. Ensemble techniques that balance high accuracy and computational economy, such as XGBoost, demonstrated higher results. Preprocessing methods that improved model performance and generalization included data augmentation, normalization, and cross-validation. The results focused on the importance of resource optimization and the likelihood of learning transfer in picture classification, contributing to an understanding of evidence for selecting the appropriate algorithms for application diversity in healthcare, agriculture, and independent systems. Future research suggested investigating amalgamation techniques and improved model interpretability for further circumstances.