The advancement of current technology has led to the widespread utilization of pattern recognition in diverse fields, such as identifying signature patterns, fingerprints, faces, and handwriting. Human handwriting exhibits variations from one person to another, often making it challenging to read or recognize, which can hinder daily activities, particularly in transactions requiring handwritten input. Handwriting, being a distinct expression of individuals, can be effectively distinguished or recognized using pattern recognition methods, particularly through computer-based classification techniques, including deep learning. In this research, deep learning was employed for the classification of handwritten characters, encompassing a total of 5200 data samples, consisting of lowercase and uppercase letters from 'a' to 'z,' with each letter represented by 100 data samples. The data underwent several stages, including pre-processing, feature extraction, classification, and evaluation. The evaluation phase employed k-fold cross-validation repeated ten times, which is a statistical technique aimed at assessing classifier performance. The study revealed that the highest accuracy, at 58.36%, was achieved using a 2-layer architecture with 512 and 256 units, while the lowest accuracy, at 43.42%, was obtained with a 5-layer architecture comprising 512, 256, 128, 64, and 32 units.