# Model Distillation for Revenue Optimization: Interpretable Personalized Pricing

@inproceedings{Biggs2021ModelDF, title={Model Distillation for Revenue Optimization: Interpretable Personalized Pricing}, author={Max Biggs and Wei Sun and Markus Ettl}, booktitle={ICML}, year={2021} }

Data-driven pricing strategies are becoming increasingly common, where customers are offered a personalized price based on features that are predictive of their valuation of a product. It is desirable to have this pricing policy be simple and interpretable, so it can be verified, checked for fairness, and easily implemented. However, efforts to incorporate machine learning into a pricing framework often lead to complex pricing policies which are not interpretable, resulting in mixed results in…

## 11 Citations

Optimal Policy Trees

- Computer ScienceMachine Learning
- 2022

We propose an approach for learning optimal tree-based prescription policies directly from data, combining methods for counterfactual estimation from the causal inference literature with recent…

Convex Loss Functions for Contextual Pricing with Observational Posted-Price Data

- Computer ScienceArXiv
- 2022

This work introduces suitable loss functions for this pricing setting which can be directly optimized to find an effective pricing policy with expected revenue guarantees without the need for estimation of an intermediate demand function, and proposes generalized hinge and quantile pricing loss functions.

Decentralized Online Convex Optimization in Networked Systems

- Computer ScienceICML
- 2022

This work proposes a novel online algorithm, Localized Predictive Control (LPC), which generalizes predictive control to multi-agent systems and achieves a competitive ratio of 1 + O ( ρ kT ) + ˜ O ( rS ) in an adversarial setting, which is the first competitive ratio bound on decentralized predictive control for networked online convex optimization.

Implementing Reinforcement Learning Datacenter Congestion Control in NVIDIA NICs

- Computer ScienceArXiv
- 2022

This work presents a Reinforcement Learning (RL) based CC solution that learns from certain traffic scenarios and successfully generalizes to others, and distill the RL neural network policy into binary decision trees to achieve the desired 𝜇 sec decision latency required for real-time inference with RDMA.

Constrained Prescriptive Trees via Column Generation

- Computer ScienceAAAI
- 2022

A novel path-based mixed-integer program (MIP) formulation is introduced which identifies a (near) optimal policy efficiently via column generation and can be represented as a multiway-split tree which is more interpretable and informative than binary-split trees due to its shorter rules.

When Personalization Harms: Reconsidering the Use of Group Attributes in Prediction

- Computer Science, PsychologyArXiv
- 2022

The results highlight the prevalence of fair use violations, demonstrate actionable interventions to mitigate harm, and underscore the need to measure the gains of personalization for all groups who provide personal data.

Regulatory Instruments for Fair Personalized Pricing

- EconomicsWWW
- 2022

Personalized pricing is a business strategy to charge different prices to individual consumers based on their characteristics and behaviors. It has become common practice in many industries nowadays…

Careful! Training Relevance is Real

- Materials Science
- 2022

This paper aims to demonstrate the efforts towards in-situ applicability of EMMARM, which aims to provide real-time information about the physical and emotional impacts of EMT in the field of operations and information management.

Acceleration techniques for optimization over trained neural network ensembles

- Computer ScienceArXiv
- 2021

The results suggest that the optimization algorithm outperforms the adaption of an state-of-the-art approach in terms of computational time and optimality gaps.

Optimizing over an ensemble of neural networks

- Computer Science
- 2021

Exper-imental evaluations of the solution methods suggest that using ensembles of neural networks yields more stable and higher quality solutions, compared to single neural networks, and that the optimization algorithm outperforms a state-of-the-art approach in terms of computational time and optimality gaps.

## References

SHOWING 1-10 OF 56 REFERENCES

Optimizing Objective Functions Determined from Random Forests

- Computer Science
- 2017

This work models the problem of optimizing a tree-based ensemble objective with the feasible decisions lie in a polyhedral set as a Mixed Integer Linear Program (MILP) and shows it can be solved to optimality efficiently using Pareto optimal Benders cuts.

Optimal Prescriptive Trees

- Computer ScienceINFORMS Journal on Optimization
- 2019

Motivated by personalized decision making, given observational data [Formula: see text] involving features [Formula: see text], assigned treatments or prescriptions [Formula: see text], and outcomes…

Interpreting Predictive Models for Human-in-the-Loop Analytics

- Computer Science
- 2018

This work proposes a novel approach to interpreting complex, blackbox machine learning models by constructing simple decision trees that summarize their reasoning process, and proves that by generating a sufficient amount of data through the active learning strategy, the extracted decision tree converges to the exact decision tree, implying that it provably avoid overfitting.

Recursive Partitioning for Personalization using Observational Data

- Computer ScienceICML
- 2017

This work proposes a new approach based on recursively partitioning the data into regimes where different treatments are optimal, extending this approach to an optimal partitioning approach that finds a globally optimal partition, achieving a compact, interpretable, and impactful personalization model.

Recursive partitioning for heterogeneous causal effects

- MathematicsProceedings of the National Academy of Sciences
- 2016

This paper provides a data-driven approach to partition the data into subpopulations that differ in the magnitude of their treatment effects, and proposes an “honest” approach to estimation, whereby one sample is used to construct the partition and another to estimate treatment effects for each subpopulation.

The Price of Interpretability

- Computer ScienceArXiv
- 2019

This paper introduces a mathematical framework in which machine learning models are constructed in a sequence of interpretable steps and quantifies the ``price'' of interpretability, i.e., the tradeoff with predictive accuracy.

The Power and Limits of Predictive Approaches to Observational-Data-Driven Optimization

- Computer Science
- 2016

Both the power and limits of predictive approaches to observational-data-driven optimization with a particular focus on pricing are studied and a new hypothesis test for causal-effect objective optimality is developed, applying it to interest-rate-setting data.

Personalized Dynamic Pricing with Machine Learning: High-Dimensional Features and Heterogeneous Elasticity

- Computer ScienceManag. Sci.
- 2021

This paper considers a seller who can dynamically adjust the price of a product at the individual customer level, by utilizing information about customers’ characteristics encoded as a d-dimensional feature vector, and designs a near-optimal pricing policy for a semiclairvoyant seller who achieves an expected regret of order.

A General Knowledge Distillation Framework for Counterfactual Recommendation via Uniform Data

- Computer ScienceSIGIR
- 2020

This paper proposes a general knowledge distillation framework for counterfactual recommendation that enables uniform data modeling through four approaches that achieve better performance over the baseline models in terms of AUC and NLL.

Decision Trees for Decision-Making under the Predict-then-Optimize Framework

- Computer ScienceICML
- 2020

It is demonstrated on these datasets that SPOTs simultaneously provide higher quality decisions and significantly lower model complexity than other machine learning approaches (e.g., CART) trained to minimize prediction error.