Soil water content is an important parameter in making a decision to use a tractor or not. The process of measuring soil water content and levels of field capacity in conventional which takes a long time and cannot be used in real-time to measure it is a major problem in the field. Determinants of soil water content such as ambient temperature, humidity, and rainfall can be obtained easily and quickly either by using a tool or retrieving data from the nearest BMKG station. The objective of this research is to obtain the most optimal prediction model in making decisions about tractor operation in dry land. This research uses an Artificial Neural Network (ANN) in modeling predictions of tractor operation. Prediction of tractor operation is a prediction of tractor use on a certain day using input data obtained before the day of tractor use. ANN modeling uses the back-propagation supervised learning method. The best ANN model used four hidden neurons with a learning coefficient of 0.2, a momentum of 0.8 and 20,000 iterations. This model has been able to provide optimal predictions with an accuracy value of 77%. The ANN model has been successful in predicting tractor operation on dry land using the back-propagation supervised learning method.