ARMDN: Associative and Recurrent Mixture Density Networks for eRetail Demand Forecasting

TL;DR

ARMDN is a neural network architecture that effectively models complex, non-stationary demand patterns in eRetail by integrating associative factors, time-series trends, and demand variance, leading to improved forecasting accuracy.

Contribution

The paper introduces AR-MDN, a novel neural network combining feature embeddings, MLP, LSTM, and mixture density networks for demand forecasting in eRetail, handling non-stationarity and associative factors.

Findings

Significant improvement over existing methods in forecasting accuracy.

Effective modeling of associative factors and demand variance.

End-to-end trainable architecture without extra supervision.

Abstract

Accurate demand forecasts can help on-line retail organizations better plan their supply-chain processes. The challenge, however, is the large number of associative factors that result in large, non-stationary shifts in demand, which traditional time series and regression approaches fail to model. In this paper, we propose a Neural Network architecture called AR-MDN, that simultaneously models associative factors, time-series trends and the variance in the demand. We first identify several causal features and use a combination of feature embeddings, MLP and LSTM to represent them. We then model the output density as a learned mixture of Gaussian distributions. The AR-MDN can be trained end-to-end without the need for additional supervision. We experiment on a dataset of an year's worth of data over tens-of-thousands of products from Flipkart. The proposed architecture yields a significant improvement in forecasting accuracy when compared with existing alternatives.