Next Meetup

Interpretable time series classification with Decision Trees
In our upcoming session Vera Shalaeva will discuss the application of decision trees to time series classification! Abstract Classification of time series is one of the most important problems in dynamic data mining. Machine learning algorithms are widely used to accomplish this task. However, in order for a domain expert to successfully analyze obtained results, it oftentimes requires knowledge of ML techniques. In addition, many algorithms use complex data transformations, which makes it impossible to interpret the results. As a consequence, it is crucial to develop new ML algorithms which produce models that a practitioner can easily use and analyze. Generally, Decision Trees provide interpretable classifiers which can be easily visualized. In time series context, a modified version of Decision Trees, namely Temporal Decision Trees (TDT), have been used to classify temporal data. In this presentation, we introduce an extension of TDT to improve interpretability and readability of yielded trees while maintaining the classification accuracy. This extension naturally brings an additional level of computational complexity. In order to address this issue, we present an approximation algorithm leading to learn classification trees faster while keeping performance characteristics. Short Bio Vera is a final year PhD researcher at the University of Grenoble-Alps working at Machine Learning group (AMA), of the Laboratory of Informatics (LIG). In 2015, she graduated from the Data Science Master program of Ensimag, at Grenoble. Before that, she worked as a Data Analyst in Samsung, Moscow, Russia.

Cowork in Grenoble

16 Boulevard Maréchal Lyautey · Grenoble

    Past Meetups (41)