Handwritten Digits Recognition using Novel Long Short Term Memory with Enhanced F- Measures Over K-Nearest Neighbour to Improve the Accuracy

Aim: The major goal of this research is to develop a model that can recognise digits utilising Long Short-Term Memory and LSTM cells, as well as to compare F scores for optical character recognition using LSTM and KNN on the Modified National Institute of Standards and Technology dataset. Material and Methods: GPower statistical software is used to estimate the sample size, with a pre-power test of 80%. The alpha error rate, which is 0.05, is a type-I error. The dataset contains 70K handwritten digit samples, 60000 of which are utilised as training samples and the remaining 10,000 as testing samples. Results: The digits were identified using Long Short Term-Memory (LSTM) and K-Nearest Neighbour (KNN) algorithms, with LSTM achieving 99 percent accuracy with a the 2-tailed significance value is 0.000 (p<0.05) and by KNN achieving 88 percent accuracy. Conclusion: The results showed that LSTM with LSTM cells performed substantially better than KNN in optical character recognition.