IJSRP, Volume 8, Issue 6, June 2018 Edition [ISSN 2250-3153]
Andy W. Chen
In this paper, I build automated machine learning models to recognize and predict handwritten digits. The models are supervised learning models used to predict every possible pair of digits. I evaluate the models by comparing the training and testing accuracies to explore the pairs that are most and least difficult to distinguish. I find that the pair 0 and 1 has the highest accuracy and the pair 3 and 5 has the lowest accuracy.