Interpretable machine learning for demand modeling with high-dimensional data using Gradient Boosting Machines and Shapley values

Abstract

Forecasting demand and understanding sales drivers are one of the most important tasks in retail analytics. However, traditionally, linear models and/or models with a small number of predictors have been predominantly used in sales modeling. Taking into account that real-world demand is naturally determined by complex substitution and complementation patterns among a large number of interrelated SKUs, nonlinear effects of prices, promotions, seasonality, as well as many other factors, their lagged values, and interactions, a realistic model has to be able to account for all that. We propose a conceptual model for sales modeling based on standard POS data available to any retailer and generate almost 500 potentially useful predictors of a focal SKU's sales accordingly. In our comparison of three classes of models, Gradient Boosting Machines outperformed Random Forests and Elastic nets. By using interpretable machine learning methods, we came up with actionable insights related to the importance of various groups of predictors from the conceptual model, as well as demonstrated how helpful it can be for marketing managers to decompose predictions into the effects of individual regressors by using an approximation of Shapley values for feature attribution.

Data-driven detection of counterpressing in professional football

“…2 Feature influence to the counterpressing prediction based on SHAP-values method to identify the individualized feature contribution to machine learning models. This method has been effectively used in different applications (Antipov et al 2020;Meng et al 2020;Ibrahim et al 2020;Anzer and Bauer 2021).…”

Predicting and identifying factors associated with undernutrition among children under five years in Ghana using machine learning algorithms

“…Another study from Bangladesh reported the highest accuracy when using an artificial neural network [ 10 ]. We may attribute the observed superiority of the XGBoost to its ability to leverage the outputs of weak sequential decision trees, where each new tree builds on the weaknesses of the previous trees to make accurate predictions and its ability to effectively handling complex, high-dimensional data for classification [ 27 ]. The SVM models followed the XGBoost model with regards to discrimination under the ROC curves in all the three undernutrition indicators.…”

Post-script—Retail forecasting: Research and practice

“…But interpretability has been a key area in the new wave of ML research and much progress has been made in recent years. Antipov and Pokryshevskaya (2020) applied a recently developed unified approach to interpreting model predictions called Shapley Additive Explanations. They showed that the Shapley additive explanation of boosting tree predictions are very insightful, uncovering the effects of prices and promotions, as well as ideas related to stockpiling and cross-product effects.…”