time seri shapelet primit data mine lexiang ye eamonn keogh classif time seri attract great interest past decad recent empir evid strongli suggest simpl nearest neighbor algorithm difficult beat time seri problem consid good new simplic implement nearest neighbor algorithm neg consequ nearest neighbor algorithm requir store search entir dataset result time space complex limit applic resourc limit sensor mere classif accuraci gain insight data work introduc time seri primit time seri shapelet address limit inform shapelet time seri subsequ sens maxim repres class show extens empir evalu divers domain algorithm base time seri shapelet primit interpret accur significantli faster state art classifi 