The impact of data set similarity and diversity on transfer learning success in time series forecasting
This research addresses the need for structured measures to explain transfer learning outcomes in time series forecasting, which is incremental as it builds on existing benchmarks by introducing new similarity and diversity metrics.
The study tackled the problem of understanding which data characteristics lead to transfer learning success in time series forecasting, finding that source-target similarity reduces forecasting bias, while source diversity improves accuracy and uncertainty estimation but increases bias.
Pre-trained models have become pivotal in enhancing the efficiency and accuracy of time series forecasting on target data sets by leveraging transfer learning. While benchmarks validate the performance of model generalization on various target data sets, there is no structured research providing similarity and diversity measures to explain which characteristics of source and target data lead to transfer learning success. Our study pioneers in systematically evaluating the impact of source-target similarity and source diversity on zero-shot and fine-tuned forecasting outcomes in terms of accuracy, bias, and uncertainty estimation. We investigate these dynamics using pre-trained neural networks across five public source datasets, applied to forecasting five target data sets, including real-world wholesales data. We identify two feature-based similarity and diversity measures, finding that source-target similarity reduces forecasting bias, while source diversity improves forecasting accuracy and uncertainty estimation, but increases the bias.