Performance Evaluation of Machine Learning and Deep Learning Models for Temperature Prediction in Poultry Farming

Abstract

Temperature is a commonly used environmental factor that directly impacts both health of chicks and production in poultry farming. The cold weather makes the environment more conductive for certain infection diseases like Newcastle and Avian influenza, whereas heat stress or high temperature cause poor food efficiency and decreased production. Hence, the prediction of temperature with the support of machine learning (ML) and deep learning (DL) models in advance is advantageous for poultry farming. Real-time temperature data is captured with the support of Internet of Things (IoT) nodes and sent to clouds by wireless communication; this data is analyzed using various machine learning models on clouds, and decisions are made based on the knowledge extracted from received data. In this article, different machine learning (Random Forest, Linear Regression) and deep learning models (LSTM, BiLSTM) are used to process the temperature and provide temperature predictions after every 10 mins. All the models are compared on the various performance evaluation factors like mean absolute error (MAE), root mean square error (RMSE), determination coefficient (R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ), and Spearman's correlation coefficient (SPCC). The comparison results show that the R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value for random forest i.e., 0.992 is highest compared to other models. Such models with high prediction rates significantly impact the environment management decisions and production on the farm.