 This study proposes a new time series forecasting model based on multi-layer of long short-term memory, MLSTM. It was evaluated against three other popular forecasting methods, bidirectional LSTM, linear regression, and LSTM and found to be superior in terms of accuracy and precision. The MLSTM model achieved a decrease in mean absolute percentage error, mape of 20%, root mean square error, RMSE of 25%, and mean absolute percentage error, mape of 10% when compared to LSTM. Additionally, it outperformed bidirectional LSTM with a decrease in mape of 10%, RMSE of 20%, and mape of 18%. Finally, it also surpassed linear regression with a decrease in mape of 2%, RMSE of 7%, and mape of 10%. These results demonstrate the superiority of the proposed MLSTM model over existing methods in terms of accuracy and precision. This article was authored by Tan Naqdin, Gokul Siddharthirin Vukharsu, Mehdi Said Mahmoudian, and others. We're article.tv, links in the description below.