Anticancer peptides (ACPs) are promising candidates for cancer treatment due to their safety and selectivity compared to conventional chemotherapies. However, identifying new ACPs through traditional experimental methods is expensive, labor-intensive, and time-consuming. With the vast amount of peptide sequence data available in the post-genomic era, there is a growing need for computational tools to expedite ACP prediction. While some statistical models have been proposed, they often impose limitations on feature selection and machine learning algorithms, reducing their efficiency and robustness. This study used the BLOSUM62 feature encoding method to create and test several classifiers for predicting ACP. The models demonstrated strong performance, achieving accuracy, precision, recall, and F1-scores of up to 92.09% on the test set. Gradient Boosting emerged as the top performer among the classifiers, while Random Forest and Support Vector Machines also delivered competitive results. These findings underscore our approach’s superior performance and robustness compared to traditional methods. Notably, BLOSUM62 played a critical role in enhancing predictive power, highlighting its effectiveness in identifying novel ACPs. This model represents a valuable tool for large-scale ACP discovery, offering a more efficient and reliable alternative to conventional techniques.