Using syntactic similarity to shorten the training time of deep learning models using time series datasets: A case study