“One of the important issues in the field of data mining is time series prediction, which has attracted the attention of many researchers in recent years.” Usually, the time series data are related to time and have specific patterns that by knowing these patterns and using them, it is possible to predict future events. One of the most common deep neural networks for predicting time series is the long-short-term memory (LSTM) neural network. The LSTM network is a recurrent neural network that can retain complex time-related information and understand complex patterns in time series. One of the important challenges of neural networks in predicting time series is determining the correct value of hyperparameters, which has a great impact on their efficiency and performance. Various hyperparameters can be considered. The hyperparameters considered in this paper are: the number of neurons in the hidden layer, the sparsity parameter, the weight reduction parameter in the cost function, and the sparsity penalty weight. The Bayesian optimization method is used to set the correct value of the hyperparameters to achieve the highest accuracy in the problem. Bayesian optimization method determines probability distribution for hyperparameters using probability theory. In order to evaluate the performance of the proposed method, a random data set has been used. The proposed method is applicable and generalizable to real data sets such as weather, environment and energy data. The obtained results show the efficiency and proper performance of the proposed method.