The goal of this study is to create a model that can recognize digits using Novel Recurrent Neural Networks (RNN) with LSTM cells and provide an F score comparison for optical character recognition versus Support Vector Machines (SVM) with Linear Kernel on the MNIST dataset. The sample estimation is done using the GPower statistical software with a pre-power test of 80%. The type-I error rate (alpha error rate) of 0.05 is considered. The dataset has 70K samples of handwritten digits, of which 60K are used as training samples and the remaining 10,000 are used as testing samples. In this research work, the digits are classified using RNN and linear SVM algorithms. The RNN attains an accuracy of 99% with a significance value of 0.171 (p 0.05), whereas the Linear SVM attains an accuracy of 87.75%. The results proved that optical character recognition using Novel RNN with LSTM cells performed much better than Linear SVM.