Exploratory Analysis in Time-Varying Data Sets: a Healthcare Network Application

Abstract

We introduce a new method for exploratory analysis of large data sets with time-varying features, where the aim is to automatically discover novel relationships between features (over some time period) that are predictive of any of a number of time-varying outcomes (over some other time period). Using a genetic algorithm, we co-evolve (i) a subset of predictive features, (ii) which attribute will be predicted (iii) the time period over which to assess the predictive features, and (iv) the time period over which to assess the predicted attribute. After validating the method on 15 synthetic test problems, we used the approach for exploratory analysis of a large healthcare network data set. We discovered a strong association, with 100% sensitivity, between hospital participation in multi-institutional quality improvement collaboratives during or before 2002, and changes in the risk-adjusted rates of mortality and morbidity observed after a 1-2 year lag. The proposed approach is a potentially powerful and general tool for exploratory analysis of a wide range of time-series data sets.

Keywords: Artificial intelligence; genetic algorithm; knowledge discovery; pattern recognition.

Fig. 1

a) Hypothesis of causality. b) Inverted hypothesis tested by the classifier.

Fig. 2

Information is extracted and aggregated from the time-series data relative to a dividing year (2004, in this example) and lag (2 years, in this example).

Fig. 3

Overall architecture of the approach, illustrated for use with the VON data set. Items outlined in red are co-evolved by GAMET.

Fig. 4

Experimental results on the VON data set. The bars indicate the frequency with which each of the individual features was selected in 10 GAMET trials. The red asterisks near the top indicate the features selected in the single best individual.