Abstract
Objective: This paper has been prepared as an effort to reassess the research studies on the relevance of Long-short term memory (LSTM) in the correction and prediction of sensor data for temperature. Methods: This methods are a multiple input parameters model of LSTM and an input parameter model of LSTM, the first one corrects one of those parameter and the next one predict the remainiiig data of one parameter of sensor data after training history data. Findings: The training data reach 87,600 with 8 parameter, the error of first model reduce to 0.13%. The training data reach 11,682, the error between real data and predicted data is from 3.4% to 0.03% Application / Improvement: The methods will be used to correct the history data of sensors and determine which seed to plant and whether to irrigate. In future, we will integrate Markov chain with LSTM to improve the precision of LSTM and reduce the size of training data.