Due to the importance of the steel industry in the national economy and the inherent complexity of operating a blast furnace, it is necessary to study ways to optimize its operation and the consumption of resources, therefore, this work aims to investigate the use of the LSTM (Long Short Term Memory) neural network to perform the prediction of the next temperature of the hot metal being produced. In this way, it is possible to support the work of the blast furnace operators, in order to optimize the consumption of resources to keep the blast furnace operating. With the results obtained from the experiments using the blast furnace operating data as a time series, it is concluded that the use of LSTM is satisfactory and that improvement of these experiments will meet the needs of the steel industry. The best result for LSTM, using 2 layers and 2048 neurons, achieved a Root Mean Square Error of 11.82ºC.