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

arXiv: https://arxiv.org/abs/1803.03800 · HTML: https://mlsanity.com/paper/1803.03800/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 10 Mar 2018.

## Built on by (most cited first, 9 of 9)

- [Criteria for Classifying Forecasting Methods](https://mlsanity.com/paper/2212.03523/) (2022)
- [Simultaneously Reconciled Quantile Forecasting of Hierarchically Related Time Series](https://mlsanity.com/paper/2102.12612/) (2021)
- [SutraNets: Sub-series Autoregressive Networks for Long-Sequence, Probabilistic Forecasting](https://mlsanity.com/paper/2312.14880/) (2023)
- [Adaptive Sampling for Probabilistic Forecasting under Distribution Shift](https://mlsanity.com/paper/2302.11870/) (2023)
- [Neural Multi-Quantile Forecasting for Optimal Inventory Management](https://mlsanity.com/paper/2112.05673/) (2021)
- [C2FAR: Coarse-to-Fine Autoregressive Networks for Precise Probabilistic Forecasting](https://mlsanity.com/paper/2312.15002/) (2023)
- [Coherent Probabilistic Aggregate Queries on Long-horizon Forecasts](https://mlsanity.com/paper/2111.03394/) (2021)
- [A Hybrid Statistical-Machine Learning Approach for Analysing Online Customer Behavior: An Empirical Study](https://mlsanity.com/paper/2212.02255/) (2022)
- [Stabilizing distribution-free probabilistic forecasts](https://mlsanity.com/paper/2605.28531/) (2026)

## Abstract (arXiv)

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.