Balanced Policy Evaluation and Learning

نویسنده

  • Nathan Kallus
چکیده

We present a new approach to the problems of evaluating and learning personalizeddecision policies from observational data of past contexts, decisions, and outcomes. Onlythe outcome of the enacted decision is available and the historical policy is unknown.These problems arise in personalized medicine using electronic health records and ininternet advertising. Existing approaches use inverse propensity weighting (or, doublyrobust versions) to make historical outcome (or, residual) data look like it were generatedby a new policy being evaluated or learned. But this relies on a plug-in approach thatrejects data points with a decision that disagrees with the new policy, leading to highvariance estimates and ineffective learning. We propose a new, balance-based approachthat too makes the data look like the new policy but does so directly by finding weightsthat optimize for balance between the weighted data and the target policy in the given,finite sample, which is equivalent to minimizing worst-case or posterior conditional meansquare error. Our policy learner proceeds as a two-level optimization problem overpolicies and weights. We demonstrate that this approach markedly outperforms existingones both in evaluation and learning, which is unsurprising given the wider support ofbalance-based weights. We establish extensive theoretical consistency guarantees andregret bounds that support this empirical success. 1arXiv:1705.07384v1[stat.ML]21May2017

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Evaluating the Health Reform Plan in Ayatollah Taleghani Hospital, Tehran in 2010- 2017 Using the Balanced Scorecard

Evaluating the Health Reform Plan in Ayatollah Taleghani Hospital, Tehran in 2010- 2017 Using the Balanced Scorecard Azar Ebrahimpour1, Khalil Alimohammadzadeh2, 3, *, Ali Maher4,5 1MSc, School of Management and Social Sciences, North Tehran Branch, Islamic Azad University, Tehran, Iran 2Associate Professor, Department of Health Services Management, School of Management and Social Sciences, N...

متن کامل

Comprehensive Performance Evaluation Using FAHP-FVIKOR Approach Based on Balanced Scorecard (BSC): A Case of Yazd's Ceramic and Tile Industry

The comprehensive assessment model considers the financial and non-financial aspects as balanced scorecard. Also, for more comprehensive results of the investigation, four aspects of the balanced scorecard were used and then by using fuzzy Delphi technique and experts' opinions, the standard aspects of each were achieved in two stages. To determine the importance of each aspect of BSC, the proce...

متن کامل

BANKS' PERFORMANCE EVALUATION MODEL BASED ON THE BALANCED SCORE CARD APPROACH, FUZZY DEMATEL AND ANALYTIC NETWORK PROCESS

<span style="color: #000000; font-family: Tahoma, sans-serif; font-size: 13px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: auto; text-align: justify; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px; display: inline !important; float: none; backgro...

متن کامل

BANKS' PERFORMANCE EVALUATION MODEL BASED ON THE BALANCED SCORE CARD APPROACH, FUZZY DEMATEL AND ANALYTIC NETWORK PROCESS

<span style="color: #000000; font-family: Tahoma, sans-serif; font-size: 13px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: auto; text-align: justify; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px; display: inline !important; float: none; backgro...

متن کامل

Evaluation of Effectiveness of Implementing Quality Management System (ISO9001:2000) Using BSC Approach in NIGC

The purpose of this research is to evaluate the effectiveness of implementation Quality Management System (ISO 9001:2000) in gas processing plants of National Iranian Gas Company using balanced scorecard approach. More specifically, we chose the balanced scorecard framework to examine five well-known gas processing plants in National Iranian Gas Company from four different perspectives in order...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • CoRR

دوره abs/1705.07384  شماره 

صفحات  -

تاریخ انتشار 2017