Deep Learning Based Prediction On Crop Yield Combined Autoencoder Long-Short Term Memory (Lstm) And Temporal Convolutional Network (Tcn)

Abstract

Population growth rate is experiencing unprecedently upward trend leading human to growing demand on food. While advancement in agricultural technology has been made, an important point is that we have limited natural resources, including water, soil nutrition, etc. Therefore, approaches to predicting crop yield should bring significant benefits of resources management, financial decisions, food supplement planning, soil health improvement and timely warning for any food chain disruptions. Machine learning, a subset of Artificial Intelligence (AI) that focuses on learning, is a useful method that can estimate yields more accurately utilizing a variety of characteristics. Random Forest (RF), Artificial Neural Network (ANN) are some of popular machine learning methos in the area. Machine learning (ML) tries to buid prediction model by determining patterns and correlation of dataset and suggesting outputs based on historical experience and its reaction with parameters. However, there are still desires to build more advance and effective models. Deep learning (DL) is branch of machine learning which researchers have found interested. It comprises multiple hidden layers of artificial neural networks and tends to give better accuracy in crop yield prediction. In recent years, combination of deep learning algorithms have drawn much attention from scholars thanks to its excellent performance. However, there are little research focused on combination of Autoencoder Long-Short Term Memory (LSTM) and Temporal Convolutional Network (TCN). In this study, we will try to examine the effectiveness in predicting crop yield.