A Novel Time Series Representation Approach for Dimensionality Reduction

With the growth of streaming data from many domains such as transportation, finance, weather, etc, there has been a surge in interest time series mining. this and massive amounts data, representation become essential for reducing dimensionality to overcome available memory constraints. Moreover, mining processes include similarity search learning historical tasks. These tasks require high computation time, which can be reduced by dimensionality. This paper proposes novel called Adaptive Simulated Annealing Representation (ASAR). ASAR considers an optimization problem with objective preserving shape looks instances raw that represent local trends neglect rest. The algorithm is adapted fulfill mentioned above. We compare three well-known approaches literature. experimental results have shown achieved highest reduction dimensions. it using representation, process accelerated most. also tested terms information performing One Nearest Neighbor (1-NN) classification K-means clustering, assures its ability preserve them outperforming competing task achieving close accuracy 1-NN task.