• Alaparthi Bhargavi
• Gopu Prasannanjali
• Chappidi Anjani Kumar

The prominent use of machine learning in image processing is handwritten digit recognition. In this study, we compare the Radial Basis Function Neural Network with Principle Component Analysis (RBFNN-PCA) and the Backpropagation Neural Network with Principal Component Analysis, two widely used methods for handwritten digit recognition (BPNNPCA). Both methods use PCA to reduce the dimensionality of the input data before using the appropriate neural network model for classification. We trained and evaluated both models using the MNIST dataset, which contains 60,000 training images and 10,000 testing 28 by 28-pixel images of handwritten digits. Our experiment results show demonstrated the accuracy of the RBFNN-PCA method exceeded the BPNN-PCA method. In particular, the RBFNNPCA’s accuracy was 98.65%, compared to the BPNN-PCA’s accuracy of 97.46%. Furthermore, we performed sensitivity analysis on the number of principal components used in both approaches. We found that the RBFNN-PCA and BPNN-PCA are more robust to changes in the number of principal components, with a minimum of 40 principal components required to achieve high accuracy. The overall effectiveness of the RBFNN-PCA approach for handwritten digit recognition is shown by our experimental results. In addition to being more precise, the RBFNN-PCA method is also more resistant to variations in the number of principal components applied to reduce dimensionality. This paper contributes to the ongoing research in this field by offering helpful insights into the choice of machine-learning learning models for handwritten digit recognition.

Handwritten digit recognition, Radial Basis Function Neural Network (RBFNN), Backpropagation Neural Network (BPNN), Principal Component Analysis (PCA), MNIST dataset.

"Handwritten Digit Recognition using RBFNN-PCA", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.10, Issue 3, page no.g523-g527, March-2023, Available :http://www.jetir.org/papers/JETIR2303680.pdf

"Handwritten Digit Recognition using RBFNN-PCA", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.10, Issue 3, page no. ppg523-g527, March-2023, Available at : http://www.jetir.org/papers/JETIR2303680.pdf