Efficient Observation Time Window Segmentation for Administrative Data Machine Learning
This addresses efficiency challenges in temporal modeling for administrative data applications, though it appears incremental as it optimizes an existing hyperparameter tuning process rather than introducing a fundamentally new approach.
The paper tackles the problem of exponential growth in hyperparameter search space when using different time resolutions for different features in administrative data machine learning, proposing a computationally efficient TAIB technique that identifies which features benefit most from time bin tuning. Results show TAIB produces models that train more efficiently and perform better than uniform binning approaches on hospital and housing/homelessness datasets.
Machine learning models benefit when allowed to learn from temporal trends in time-stamped administrative data. These trends can be represented by dividing a model's observation window into time segments or bins. Model training time and performance can be improved by representing each feature with a different time resolution. However, this causes the time bin size hyperparameter search space to grow exponentially with the number of features. The contribution of this paper is to propose a computationally efficient time series analysis to investigate binning (TAIB) technique that determines which subset of data features benefit the most from time bin size hyperparameter tuning. This technique is demonstrated using hospital and housing/homelessness administrative data sets. The results show that TAIB leads to models that are not only more efficient to train but can perform better than models that default to representing all features with the same time bin size.