A Similarity-based Time Series Source Dataset Selection Method for Transfer Learning

Abstract Source datasets are selected with high similarity to target improve the effect of transfer learning. The low between source and may lead negative transfer. This paper proposes a similarity-based time series dataset selection method for First, reduce complexity operation, we use Dynamic time-warped barycentric averaging obtain prototype signals each convert calculation into signals. dynamic warping (DTW) algorithm, commonly used calculate time-series signal similarity, does not consider amount data contains on calculation. article makes improvements above issues. Experiments carried out 20-time archived by University California Riverside. Experimental results show that average proposed is better than measurement based DTW.