Active Learning for Decision-Making from Imbalanced Observational Data

Iiris Sundin, Peter Schulam, Eero Siivola, Aki Vehtari, Suchi Saria, Samuel Kaski
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:6046-6055, 2019.

Abstract

Machine learning can help personalized decision support by learning models to predict individual treatment effects (ITE). This work studies the reliability of prediction-based decision-making in a task of deciding which action $a$ to take for a target unit after observing its covariates $\tilde{x}$ and predicted outcomes $\hat{p}(\tilde{y} \mid \tilde{x}, a)$. An example case is personalized medicine and the decision of which treatment to give to a patient. A common problem when learning these models from observational data is imbalance, that is, difference in treated/control covariate distributions, which is known to increase the upper bound of the expected ITE estimation error. We propose to assess the decision-making reliability by estimating the ITE model’s Type S error rate, which is the probability of the model inferring the sign of the treatment effect wrong. Furthermore, we use the estimated reliability as a criterion for active learning, in order to collect new (possibly expensive) observations, instead of making a forced choice based on unreliable predictions. We demonstrate the effectiveness of this decision-making aware active learning in two decision-making tasks: in simulated data with binary outcomes and in a medical dataset with synthetic and continuous treatment outcomes.

Cite this Paper


BibTeX
@InProceedings{pmlr-v97-sundin19a, title = {Active Learning for Decision-Making from Imbalanced Observational Data}, author = {Sundin, Iiris and Schulam, Peter and Siivola, Eero and Vehtari, Aki and Saria, Suchi and Kaski, Samuel}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {6046--6055}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/sundin19a/sundin19a.pdf}, url = {https://proceedings.mlr.press/v97/sundin19a.html}, abstract = {Machine learning can help personalized decision support by learning models to predict individual treatment effects (ITE). This work studies the reliability of prediction-based decision-making in a task of deciding which action $a$ to take for a target unit after observing its covariates $\tilde{x}$ and predicted outcomes $\hat{p}(\tilde{y} \mid \tilde{x}, a)$. An example case is personalized medicine and the decision of which treatment to give to a patient. A common problem when learning these models from observational data is imbalance, that is, difference in treated/control covariate distributions, which is known to increase the upper bound of the expected ITE estimation error. We propose to assess the decision-making reliability by estimating the ITE model’s Type S error rate, which is the probability of the model inferring the sign of the treatment effect wrong. Furthermore, we use the estimated reliability as a criterion for active learning, in order to collect new (possibly expensive) observations, instead of making a forced choice based on unreliable predictions. We demonstrate the effectiveness of this decision-making aware active learning in two decision-making tasks: in simulated data with binary outcomes and in a medical dataset with synthetic and continuous treatment outcomes.} }
Endnote
%0 Conference Paper %T Active Learning for Decision-Making from Imbalanced Observational Data %A Iiris Sundin %A Peter Schulam %A Eero Siivola %A Aki Vehtari %A Suchi Saria %A Samuel Kaski %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-sundin19a %I PMLR %P 6046--6055 %U https://proceedings.mlr.press/v97/sundin19a.html %V 97 %X Machine learning can help personalized decision support by learning models to predict individual treatment effects (ITE). This work studies the reliability of prediction-based decision-making in a task of deciding which action $a$ to take for a target unit after observing its covariates $\tilde{x}$ and predicted outcomes $\hat{p}(\tilde{y} \mid \tilde{x}, a)$. An example case is personalized medicine and the decision of which treatment to give to a patient. A common problem when learning these models from observational data is imbalance, that is, difference in treated/control covariate distributions, which is known to increase the upper bound of the expected ITE estimation error. We propose to assess the decision-making reliability by estimating the ITE model’s Type S error rate, which is the probability of the model inferring the sign of the treatment effect wrong. Furthermore, we use the estimated reliability as a criterion for active learning, in order to collect new (possibly expensive) observations, instead of making a forced choice based on unreliable predictions. We demonstrate the effectiveness of this decision-making aware active learning in two decision-making tasks: in simulated data with binary outcomes and in a medical dataset with synthetic and continuous treatment outcomes.
APA
Sundin, I., Schulam, P., Siivola, E., Vehtari, A., Saria, S. & Kaski, S.. (2019). Active Learning for Decision-Making from Imbalanced Observational Data. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:6046-6055 Available from https://proceedings.mlr.press/v97/sundin19a.html.

Related Material