• shapelets
• xai
• time series
• explainability
• shap
• multivariate time series

After analyzing a neural network used for detecting car accidents, more efficient and interpretable alternatives are proposed. Subsequently, an innovative method based on Multivariate Shapelets is introduced, which can achieve better results than the original neural network efficiently and can be easily interpreted by a third-party operator. Finally, this new method is tested on other datasets to evaluate its performance.