This research explores the application of Optical Character Recognition (OCR) in the realm of computer science, delving into the intersection of image processing, machine learning, and neural networks. The project comprises two main components: the training phase and the testing phase. In the training phase, a novel algorithm is employed to teach a neural network to recognize diverse characters by exposing it to various sets of similar but not identical characters. The testing phase involves evaluating the neural network's performance on a new dataset, akin to assessing a trained individual's ability to recognize characters. The project incorporates statistical modeling and optimization techniques, emphasizing the significance of statistical concepts, optimizer techniques, and filtering processes. Mathematical and predictive aspects underpin the algorithms, guiding the creation of a machine learning model. The research underscores the intricate interplay between prediction, programming, and the implementation of neural networks for character recognition.