The objective of this research project is to devise a Multilevel Adaptive Reduction Technique (MART) for time series data. MART will extract the main features of a time series and encode them in a reduced data format called reduct. The extracted features will be leveraged to establish a distance between reducts that satisfies the (Euclidean distance) lower bounding constraint. Furthermore, we will strive to make MART applicable on the reduced data and at different levels. We will conduct experiments on time series datasets that show MART accuracy and multilevel reduction powers.