The identification of oreochromis niloticus feeding behaviour through the integration of photoelectric sensor and logistic regression classifier

• Main Authors: Mohamad Radzi, Mohd Sojak, Mohd Azraai, Mohd Razman, Anwar, P. P. Abdul Majeed, Rabiu Muazu, Musa, Ahmad Shahrizan, Abdul Ghani, Ismed, Iskandar
• Format: Conference or Workshop Item
• Language: English; English
• Published: Universiti Malaysia Pahang 2018
• Subjects: TS Manufactures
• Online Access: http://umpir.ump.edu.my/id/eprint/24587/1/35.%20The%20identification%20of%20oreochromis%20niloticus%20feeding%20behaviour.pdf; http://umpir.ump.edu.my/id/eprint/24587/2/35.1%20The%20identification%20of%20oreochromis%20niloticus%20feeding%20behaviour.pdf; http://umpir.ump.edu.my/id/eprint/24587/

Oreochromis niloticus or tilapia is the second major freshwater aqua- culture bred after catfish in Malaysia. By understanding the feeding behaviour, fish farmers will able to identify the best feeding routine. In the present investi- gation, photoelectric sensors are used to identify the movement, speed and posi- tion of the fish. The signals acquired from the sensors are converted into binary data. The hunger behaviour classes are determined through k-means clustering algorithm, i.e., satiated and unsatiated. The Logistic Regression (LR) classifier was employed to classify the aforesaid hunger state. The model was trained by means of 5-fold cross-validation technique. It was shown that the LR model is able to yield a classification accuracy for tested data during the day at three dif- ferent time windows (4 hours each) is 100%, 88.7% and 100%, respectively, whilst the for-night data it was shown to demonstrate 100% classification accu- racy.