Simulation for time series classification using feature covariance matrices with K-nearest neighbor

Covariance matrices have advantages to classify series for time classification. The is segmented into subsequences that its feature points are adopted as the vector. can capture pairwise correlation between vectors. After extraction, next step classification process. Our primary purpose evaluate covariance with kNN (Cov-kNN) under simulation pattern and UCR data sets by optimal neighborhood interval. CovNN also compared in this work. results both algorithms comparable performance. So that, using Cov-kNN separating several intervals considered.