Title: PigB: intelligent pig breeds classification using supervised machine learning algorithms

Authors: Pritam Ghosh; Satyendra Nath Mandal

Addresses: Department of Computer Science and Technology, Iswar Chandra Vidyasagar Polytechnic, Sevayatan, Jhargram, West Bengal, 721514, India ' Department of Information Technology, Kalyani Government Engineering College, Kalyani, Nadia, West Bengal, 741235, India

Abstract: Different supervised machine learning algorithms' performance varies when applied to different datasets. Moreover, using a generalised supervised algorithm may not be able to produce the optimal performance as the nature of data is different for different datasets. In this paper, we used a pig breed dataset containing various statistical features extracted from individual pig images of five pig breeds. Eight well-established algorithms such as logistic regression, multilayer perceptron, decision trees, gradient boosted decision trees, random forest, support vector machine, K-nearest neighbours, and naïve Bayes are carefully applied to this dataset by tuning the necessary hyperparameters for each algorithm. The performance of all the applied algorithms is compared based on the micro averaged area under precision-recall (PR) and receiver operator characteristics (ROC) curves. From the results obtained, the SVM algorithm with a radial basis function (RBF) kernel has outperformed all the other algorithms with a pig breed prediction accuracy of 98%.

Keywords: supervised learning; fine grained classification; statistical features; performance evaluation; hyperparameter optimisation; receiver operator characteristics; ROC; support vector machine; SVM; radial basis function; RBF.

DOI: 10.1504/IJAISC.2022.126345

International Journal of Artificial Intelligence and Soft Computing, 2022 Vol.7 No.3, pp.242 - 266

Received: 02 Jul 2022
Received in revised form: 09 Aug 2022
Accepted: 09 Aug 2022
Published online: 21 Oct 2022 *