A Comparative Overview of Classification Algorithm for Bangla Handwritten Digit Recognition
| Autor: | Mohammad Mohaiminul Islam, Md. Nazmul Hoq, Nadira Anjum Nipa, Md. Mostofa Akbar |
|---|---|
| Rok vydání: | 2019 |
| Předmět: | |
| Zdroj: | Proceedings of International Joint Conference on Computational Intelligence ISBN: 9789811375637 IJCCI |
| Popis: | Bangla handwritten digit recognition (BHDR) is a well-known problem in the digitization of Bangla language and Bengali script. A lot of work has been done on BHDR and very good accuracy has been achieved. This success can be extended to handwritten Bangla character (vowel, consonant) recognition which will result in automatic understanding of Bangla handwritings. But the main difficulty is faced when it comes to choosing an appropriate classification algorithm to recognize the character of Bengali handwritten script. In this paper, a comparative overview of classification algorithms for BHDR has been provided which will make it easy to decide an appropriate classification algorithm. Here, we have shown a broad comparison of eight (08) different classification algorithms using CMATERdb 3.1.1 Bangla Handwritten Numeral datasets. Different evaluation metrics have been used to justify the comparative analysis. Artificial Neural Network (ANN) performed best whereas Logistic Regression performed well compared to others in terms of the sensitivity, specificity, and error rate. This comparative overview will help scientist especially new researcher to give a quick start with Bangla handwritten character recognition and digitalization. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|