# The Effectiveness of Discretization in Forecasting: An Empirical Study on Neural Time Series Models

arXiv: https://arxiv.org/abs/2005.10111 · HTML: https://mlsanity.com/paper/2005.10111/

## Key facts

- Cited by 18 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 20 May 2020.

## Built on by (most cited first, 18 of 18)

- [Chronos: Learning the Language of Time Series](https://mlsanity.com/paper/2403.07815/) (2024)
- [Large Language Models for Time Series: A Survey](https://mlsanity.com/paper/2402.01801/) (2024)
- [The Capacity and Robustness Trade-off: Revisiting the Channel Independent Strategy for Multivariate Time Series Forecasting](https://mlsanity.com/paper/2304.05206/) (2023)
- [Hierarchical Classification Auxiliary Network for Time Series Forecasting](https://mlsanity.com/paper/2405.18975/) (2024)
- [Kronos: A Foundation Model for the Language of Financial Markets](https://mlsanity.com/paper/2508.02739/) (2025)
- [A comparison of LSTM and GRU networks for learning symbolic sequences](https://mlsanity.com/paper/2107.02248/) (2021)
- [SutraNets: Sub-series Autoregressive Networks for Long-Sequence, Probabilistic Forecasting](https://mlsanity.com/paper/2312.14880/) (2023)
- [Deep Non-Parametric Time Series Forecaster](https://mlsanity.com/paper/2312.14657/) (2023)
- [Spliced Binned-Pareto Distribution for Robust Modeling of Heavy-tailed Time Series](https://mlsanity.com/paper/2106.10952/) (2021)
- [VQ-AR: Vector Quantized Autoregressive Probabilistic Time Series Forecasting](https://mlsanity.com/paper/2205.15894/) (2022)
- [SensorBench: Benchmarking LLMs in Coding-Based Sensor Processing](https://mlsanity.com/paper/2410.10741/) (2024)
- [Multimodal Language Models with Modality-Specific Experts for Financial Forecasting from Interleaved Sequences of Text and Time Series](https://mlsanity.com/paper/2509.19628/) (2025)
- [C2FAR: Coarse-to-Fine Autoregressive Networks for Precise Probabilistic Forecasting](https://mlsanity.com/paper/2312.15002/) (2023)
- [A Comparative Study on How Data Normalization Affects Zero-Shot Generalization in Time Series Foundation Models](https://mlsanity.com/paper/2512.02833/) (2025)
- [Predicting UAV Type: An Exploration of Sampling and Data Augmentation for Time Series Classification](https://mlsanity.com/paper/2403.00565/) (2024)
- [Spatiodynamic inference using vision-based generative modelling](https://mlsanity.com/paper/2507.22256/) (2025)
- [Discrete Tokenization Unlocks Transformers for Calibrated Tabular Forecasting](https://mlsanity.com/paper/2603.07448/) (2026)
- [When Does Retrieval Help Time-Series Forecasting?](https://mlsanity.com/paper/2609.20193/) (2026)

## Abstract (arXiv)

Time series modeling techniques based on deep learning have seen many advancements in recent years, especially in data-abundant settings and with the central aim of learning global models that can extract patterns across multiple time series. While the crucial importance of appropriate data pre-processing and scaling has often been noted in prior work, most studies focus on improving model architectures. In this paper we empirically investigate the effect of data input and output transformations on the predictive performance of several neural forecasting architectures. In particular, we investigate the effectiveness of several forms of data binning, i.e. converting real-valued time series into categorical ones, when combined with feed-forward, recurrent neural networks, and convolution-based sequence models. In many non-forecasting applications where these models have been very successful, the model inputs and outputs are categorical (e.g. words from a fixed vocabulary in natural language processing applications or quantized pixel color intensities in computer vision). For forecasting applications, where the time series are typically real-valued, various ad-hoc data transformations have been proposed, but have not been systematically compared. To remedy this, we evaluate the forecasting accuracy of instances of the aforementioned model classes when combined with different types of data scaling and binning. We find that binning almost always improves performance (compared to using normalized real-valued inputs), but that the particular type of binning chosen is of lesser importance.