Neural Network Design to Determine Variables Affecting Poultry Growth

Abstract

This paper presents the prediction of poultry growth in relation to the environmental variables that most influence this process using LSTM networks. This required the installation of sensors in a warehouse of the company Nutravi located in Meta, to obtain real-time data on temperature, humidity, wind speed and weight of the chickens, where these measurements are reflected in ThingSpeak being a program focused on IoT. Subsequently, the data are used for the application of recurrent neural networks performing the prediction of the growth of the animals taking as inputs the environmental variables. With this, the definition of the network model and the parameters to be configured to ensure good learning, which is reflected in a web page designed in Python, being able to visualize the ideal weight of the bird and the weight generated by the prediction to validate that it is in the appropriate ranges of growth.