Study on feeding prediction model of scallop (Patinopecten yessoensis) seedlings based on BP neural network

To solve the problems of extensive manual feeding methods and poor accuracy in the scallop (Patinopecten yessoensis) seedlings, a feeding prediction model based on a BP neural network for the P. yessoensis seedlings was proposed. The number of days of growth, feed consumption, and water temperature of P. yessoensis seedlings were taken as input vectors of the prediction model, the feeding amount was taken as output vector, and the feeding amount was taken as output vector, and the mapping relationship between the data was mined by BP neural network to establish the prediction model, the accuracy and stability of the model were verified by feeding test. The results showed that after being automatically fed by the system for 3 d, the root mean square error of 30 d P. yessoensis seedlings decreased from 456.6 × 10⁵ for manual feeding to 226.6 × 10⁵, a reduction of 50.34%. The absolute percentage error of automatic feeding was 0.041, which was lower than that of manual feeding at 0.043; after being automatically fed by the system for 3 d, the root mean square error of the P. yessoensis seedlings cultivated for 42 d decreased from 194.2 × 10⁵ for manual feeding to 149.3 × 10⁵, a decrease of 23.09%. The absolute percentage error of automatic feeding was 0.020, which was less than 0.039 for manual feeding, which indicated that the accuracy and stability of the prediction model of feeding of P. yessoensis seedlings were better than that of manual, which provided an important reference for the research and development of automatic feeding equipment for Patinopecten yessoensis seedlings.