DataHour: Dynamic Time Warping for Time Series Classification

Online 27-08-2022 01:30 PM to 27-08-2022 02:30 PM
  • 9579


  • Knowledge and Learning


DataHour Recording

Find the resources used in the DataHour HERE.

About the DataHour:

The seasonal growth of wheat is studied for the months from October 2020 to May 2020 for the three districts in India i.e., Karnal, Kaithal and Dewas. The metric used to study the same is Vegetation Index which is normalized(termed NDVI) and precomputed from Sentinel-2 satellite data. The training data comprises the geographical information of the district i.e., latitude and longitude along with the information if wheat can be produced on it or not. It also contains the NDVI data from the date of germination till harvest for each sector in the district which is the primary key for the former and acts as a foreign key for the latter.

The NDVI time series for the districts is analyzed for similarity using Dynamic Time Warping (DTW). The DTW is used as feature embeddings and as a metric for 1NN classification.  The two approaches are applied on the test data and evaluation metrics are compared. For the feature embeddings, two classifiers i.e. Support Vector Machine and tree-based ensemble methods are used. In the case of the tree-based ensemble method, Random Forest Classifier was used. The evaluation results of the two classifiers are checked for the individual as well as combined districts and the scenarios are analyzed and discussed for cases where one classifier outperforms the other. Based on the experiments performed on the given datasets, it is concluded that using DTW as feature embeddings outperforms using DTW as a metric for predicting if wheat can be grown in the area or not.

Prerequisites:  Enthusiasm for learning Data Science and basics of NumPy, Pandas, Classification, Hyperparameter Tuning and Colab Notebook.

Who is this DataHour for?

  • Students with an interest in Data Science and Time Series.
  • Data science professionals who want to accelerate their career growth.


Sumeet Lalla

Data Scientist at Cognizant

Masters of Data Science from Higher School Of Economics Moscow and Bachelors of Engineering in Computer Engineering from Thapar University. 5.5 years of experience in Data Science and Software Engineering. Working as a Data Scientist in Cognizant and have previously worked as Software Developer in Siemens Technology And Services and Technology Analyst in Deloitte Consulting and Pvt Ltd.

You can follow him on Linkedin.


Please register/login to participate in the contest

Please register to participate in the contest

Please register to participate in the contest



We use cookies essential for this site to function well. Please click to help us improve its usefulness with additional cookies. Learn about our use of cookies in our Privacy Policy.


We believe in making Analytics Vidhya the best experience possible for Data Science enthusiasts. Help us by providing valuable Feedback.