skip to main content

Explore Research Products in the NSF-PAR

Title: Market Segmentation Trees
Problem definition: We seek to provide an interpretable framework for segmenting users in a population for personalized decision making. Methodology/results: We propose a general methodology, market segmentation trees (MSTs), for learning market segmentations explicitly driven by identifying differences in user response patterns. To demonstrate the versatility of our methodology, we design two new specialized MST algorithms: (i) choice model trees (CMTs), which can be used to predict a user’s choice amongst multiple options, and (ii) isotonic regression trees (IRTs), which can be used to solve the bid landscape forecasting problem. We provide a theoretical analysis of the asymptotic running times of our algorithmic methods, which validates their computational tractability on large data sets. We also provide a customizable, open-source code base for training MSTs in Python that uses several strategies for scalability, including parallel processing and warm starts. Finally, we assess the practical performance of MSTs on several synthetic and real-world data sets, showing that our method reliably finds market segmentations that accurately model response behavior. Managerial implications: The standard approach to conduct market segmentation for personalized decision making is to first perform market segmentation by clustering users according to similarities in their contextual features and then fit a “response model” to each segment to model how users respond to decisions. However, this approach may not be ideal if the contextual features prominent in distinguishing clusters are not key drivers of response behavior. Our approach addresses this issue by integrating market segmentation and response modeling, which consistently leads to improvements in response prediction accuracy, thereby aiding personalization. We find that such an integrated approach can be computationally tractable and effective even on large-scale data sets. Moreover, MSTs are interpretable because the market segments can easily be described by a decision tree and often require only a fraction of the number of market segments generated by traditional approaches. Disclaimer: This work was done prior to Ryan McNellis joining Amazon. Funding: This work was supported by the National Science Foundation [Grants CMMI-1763000 and CMMI-1944428]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/msom.2023.1195 .  more » « less
Award ID(s):
1763000 1944428
NSF-PAR ID:
10442643
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
Manufacturing & Service Operations Management
Volume:
25
Issue:
2
ISSN:
1523-4614
Page Range / eLocation ID:
648 to 667
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ( , Manufacturing & Service Operations Management)

    Problem definition: Delays in admission to rehabilitation care can adversely impact patient outcomes. In addition, delayed patients keep occupying their acute care beds, making them unavailable for incoming patients. Admission delays are mainly caused by a lack of rehabilitation bed capacity and the time required to plan for rehabilitation activities, which we refer to as processing times. Because of non-standard bed allocation decisions and data limitations in practice, quantifying the magnitude of the two sources of delays can be technically challenging yet critical to the design of evidence-based interventions to reduce delays. We propose an empirical approach to understanding the contributions of the two sources of delays when only a single (combined) measure of admission delay is available. Methodology/results: We propose a hidden Markov model (HMM) to estimate the unobserved processing times and the status-quo bed allocation policy. Our estimation results quantify the magnitude of processing times versus capacity-driven delays and provide insights into factors impacting the bed allocation decision. We validate our estimated policy using a queueing model of patient flow and find that ignoring processing times or using simple bed allocation policies can lead to highly inaccurate delay estimates. In contrast, our estimated policy allows for accurate evaluation of different operational interventions. We find that reducing processing times can be highly effective in reducing admission delays and bed-blocking costs. In addition, allowing early transfer—whereby patients can complete some of their processing requirements in the rehabilitation unit—can significantly reduce admission delays, with only a small increase in rehab LOS. Managerial implications: Our study demonstrates the importance of quantifying different sources of delays in the design of effective operational interventions for reducing delays in admission to rehabilitation care. The proposed estimation framework can be applied in other transition-of-care settings with personalized capacity allocation decisions and hidden processing delays.

    History: This paper was selected for Fast Track in the M&SOM journal from the 2022 MSOM Healthcare SIG Conference.

    Funding: J. Dong was supported in part by the National Science Foundation [Grant CMMI-1762544]. V. Sarhangian was supported in part by the Natural Sciences and Engineering Research Council of Canada [Grant RGPIN-2018-04518] and the Connaught Fund.

    Supplemental Material: The e-companion is available at https://doi.org/10.1287/msom.2022.0377 .

     
    more » « less
  2. ; ; ( , Management Science)
    A wide body of empirical research has revealed the descriptive shortcomings of expected value and expected utility models of risky decision making. In response, numerous models have been advanced to predict and explain people’s choices between gambles. Although some of these models have had a great impact in the behavioral, social, and management sciences, there is little consensus about which model offers the best account of choice behavior. In this paper, we conduct a large-scale comparison of 58 prominent models of risky choice, using 19 existing behavioral data sets involving more than 800 participants. This allows us to comprehensively evaluate models in terms of individual-level predictive performance across a range of different choice settings. We also identify the psychological mechanisms that lead to superior predictive performance and the properties of choice stimuli that favor certain types of models over others. Moreover, drawing on research on the wisdom of crowds, we argue that each of the existing models can be seen as an expert that provides unique forecasts in choice predictions. Consistent with this claim, we find that crowds of risky choice models perform better than individual models and thus provide a performance bound for assessing the historical accumulation of knowledge in our field. Our results suggest that each model captures unique aspects of the decision process and that existing risky choice models offer complementary rather than competing accounts of behavior. We discuss the implications of our results on theories of risky decision making and the quantitative modeling of choice behavior. This paper was accepted by Yuval Rottenstreich, behavioral economics and decision analysis. 
    more » « less
  3. ; ; ( , Proceedings of the 27th ACM Conference on User Modeling, Adaptation and Personalization)
    Identifying instances when a user will not able to attend to an incoming message and constructing an auto-response with relevant contextual information may help reduce social pressures to immediately respond that many users face. Mobile messaging behavior often varies from one person to another. As a result, compared to a generic model considering profiles of several users, a personalized model can capture a user's messaging behavior more accurately to predict their inattentive states. However, creating accurate personalized models requires a non-trivial amount of individual data, which is often not available for new users. In this work, we investigate a weighted hybrid approach to model users' attention to messaging. Through dynamic performance-based weighting, we combine the predictions of three types of models, a general model, a group model and a personalized model to create an approach which can work through the lack of initial data while adapting to the user's behavior. We present the details of our modeling approach and the evaluation of the model with over three weeks of data from 274 users. Our results highlight the value of hybrid weighted modeling to predict when a user cannot attend to their messages. 
    more » « less
  4. ; ( , JCDL '22: Proceedings of the 22nd ACM/IEEE Joint Conference on Digital Libraries)
    There is substantial evidence from behavioral economics and decision sciences demonstrating that in the context of decision-making under uncertainty, the carriers of value behind actions are gains and losses defined relative to a reference point (e.g. pre-action expectations), rather than the absolute final outcomes. Also, the capability of early predicting session-level search decisions and user experience is essential for developing reactive and proactive search recommendations. To address these research gaps, our study aims to 1) develop reference dependence features based on a series of simulated user expectations or reference points in first query segments of sessions, and 2) examine the extent to which we can enhance the performance of early predicting session behavior and user satisfaction by constructing and employing reference dependence features. Based on the experimental results on three datasets of varying types, we found that incorporating reference dependent features developed in first query segments into prediction models achieves better performance than using baseline cost-benefit features only in early predicting three key session metrics (user satisfaction score, session clicks, and session dwell time). Also, when running simulations by varying the search time expectation and rate of user satisfaction decay, the results demonstrate that users tended to expect to complete their search within a minute and showed a rapid rate of satisfaction decay in a logarithmic fashion once surpassing the estimated expectation points. By factoring in a user's search time expectation and measuring their behavioral response once the expectation is not met, we can further improve the performance of early prediction models and enhance our understanding of users' behavioral patterns. 
    more » « less
  5. ( , Proceedings of the International Joint Conference on Neural Networks)
    Human decision making behavior is observed with choice-response time data during psychological experiments. Drift-diffusion models of this data consist of a Wiener first-passage time (WFPT) distribution and are described by cognitive parameters: drift rate, boundary separation, and starting point. These estimated parameters are of interest to neuroscientists as they can be mapped to features of cognitive processes of decision making (such as speed, caution, and bias) and related to brain activity. The observed patterns of RT also reflect the variability of cognitive processes from trial to trial mediated by neural dynamics. We adapted a SincNet-based shallow neural network architecture to fit the Drift-Diffusion model using EEG signals on every experimental trial. The model consists of a SincNet layer, a depthwise spatial convolution layer, and two separate FC layers that predict drift rate and boundary for each trial in-parallel. The SincNet layer parametrized the kernels in order to directly learn the low and high cutoff frequencies of bandpass filters that are applied to the EEG data to predict drift and boundary parameters. During training, model parameters were updated by minimizing the negative log likelihood function of WFPT distribution given trial RT. We developed separate decision SincNet models for each participant performing a two-alternative forced-choice task. Our results showed that single-trial estimates of drift and boundary performed better at predicting RTs than the median estimates in both training and test data sets, suggesting that our model can successfully use EEG features to estimate meaningful single-trial Diffusion model parameters. Furthermore, the shallow SincNet architecture identified time windows of information processing related to evidence accumulation and caution and the EEG frequency bands that reflect these processes within each participant. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email