Interpretable feature learning and classification
From time series feature tweaking to temporal abstractions in medical records
Date and time: Tuesday 23 June 2020, 3pm – 4pm
Speaker: Panos Papapetrou (SU)
Abstract: The first part of the talk will tackle the issue of interpretability and explainability of opaque machine learning models, with focus on time series classification. Time series classification has received great attention over the past decade with a wide range of methods focusing on predictive performance by exploiting various types of temporal features. Nonetheless, little emphasis has been placed on interpretability and explainability. This talk will formulate the novel problem of explainable time series tweaking, where, given a time series and an opaque classifier that provides a particular classification decision for the time series, the objective is to find the minimum number of changes to be performed to the given time series so that the classifier changes its decision to another class. Moreover, it will be shown that the problem is NP-hard. Two instantiations of the problem will be presented. The classifier under investigation will be the random shapelet forest classifier. Moreover, two algorithmic solutions for the two problem instantiations will be presented along with simple optimizations, as well as a baseline solution using the nearest neighbor classifier. The second part of the talk will focus on temporal predictive models and methods for learning from sparse Electronic Health Records. The main application area is the detection of adverse drug events by exploiting temporal features and applying different levels of abstraction, without compromising predictive performance in terms of AUC.
Bio: Panagiotis Papapetrou is Professor at the Dept. of Computer and Systems Sciences at SU, with a PhD in Computer Science from Boston University. His area of expertise is algorithmic data mining with focus on temporal data mining and particular emphasis on healthcare data. He is currently PI for two related national projects (see Sec. 3). He is Action Editor of the Data Mining and Knowledge Discovery (DMKD) journal, and Guest Editor of the DMKD special issue on Mining for Health and the special issue on Big Data in Epidemics for the Big Data Research journal.