This research suggests a way to select a source dataset for a target dataset among time series datasets by measuring transferability among the datasets for better transfer learning performance. This research reviews several metrics commonly used for image classifiers to evaluate the transferability of source and target datasets. We can propose that a metric that shows correlation with accuracy is helpful for transfer learning of time series data. Finally, as a reexamination of the proposed method, we can apply the method to new datasets that are not related to research.