This book examines the impact of such constraints on elastic time-series similarity measures and provides guidance on selecting suitable measures. Time-series classification frequently relies on selecting an appropriate similarity or distance measure to compare time series effectively, often using dynamic programming techniques for more robust results. However, these techniques can be computationally demanding, which results in the usage of global constraints to reduce the search area in the dynamic programming matrix. While these constraints cut computation time significantly (by up to three orders of magnitude), they may also affect classification accuracy.



Additionally, the importance of the nearest neighbor classifier (1NN) is emphasized for its strong performance in time-series classification, alongside the kNN classifier which offers stable results. This book further explores the weighted kNN classifier, which gives closer neighbors more influence, showing how it merges accuracy and stability for improved classification outcomes.