Paper: arXiv 2311.14731

Authors: Shalini Sharma, Angshul Majumdar, Emilie Chouzenoux, Victor Elvira

Abstract

Our work presents two fundamental contributions. On the application side, we tackle the challenging problem of predicting day-ahead crypto-currency prices. On the methodological side, a new dynamical modeling approach is proposed. Our approach keeps the probabilistic formulation of the state-space model, which provides uncertainty quantification on the estimates, and the function approximation ability of deep neural networks. We call the proposed approach the deep state-space model. The experiments are carried out on established cryptocurrencies (obtained from Yahoo Finance). The goal of the work has been to predict the price for the next day. Benchmarking has been done with both state-of-the-art and classical dynamical modeling techniques. Results show that the proposed approach yields the best overall results in terms of accuracy.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper proposes a novel deep state-space model, integrating non-negative matrix factorization with Gaussian SSMs and requiring an alternating majorization-minimization algorithm for inference, demonstrating high mathematical density. Empirically, it includes benchmarks on real cryptocurrency data from Yahoo Finance and accuracy metrics, suggesting a backtest-ready implementation, though lacking publicly available code or exhaustive statistical validation.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Predict Day-Ahead<br>Cryptocurrency Prices"] --> B["Data Collection<br>Yahoo Finance<br>Eth & Btc Data"]
  B --> C["Methodology<br>Deep State-Space Model<br>SSM + DNN"]
  C --> D["Computational Process<br>Probabilistic Formulation<br>+ Uncertainty Quantification"]
  D --> E["Comparison<br>vs SOTA & Classical Models"]
  E --> F["Outcome<br>Best Overall<br>Accuracy Achieved"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a deep state-space model that combines probabilistic state-space formulation with deep neural networks for day-ahead cryptocurrency price prediction.

Source sentence
We call the proposed approach the deep state-space model.

Assumes 1

  • The EM strategy assumes non‑volatility of the parameters over the entire sequence, which is relaxed to a small window in the online implementation. model
    Source sentence
    The major drawback of the Expected-Maximization (EM) strategy used in the proposed approach is that it requires reprocessing on the entire sequence to estimate the state, control input, and observation transition operators / matrices.

Data

  • cryptocurrency dataset · Yahoo Finance market-prices public
    01/01/2014 to 01/06/2021 · ten cryptocurrencies (Litecoin, Namecoin, Dogecoin, Peercoin, Bitcoin, Ripple, NXT, Gridco
    Source sentence
    The dataset is extracted from the Yahoo finance cryptocurrency repository using Yahoo finance API 3. It consists of active cryptocurrencies ranging from some old and new cryptocurrencies. The data extracted is about eight years (between 01/01/2014 to 01/06/2021) for Litecoin, Namecoin, Dogecoin, Peercoin, Bitcoin, Ripp

Recognised sources: Yahoo Finance

Costs

Not applicable The paper reports no trading or execution result.

Finds 2

  • The proposed deep state-space model yields the best overall accuracy compared to state‑of‑the‑art and classical dynamical modeling techniques. predictive-accuracy
    Source sentence
    Results show that the proposed approach yields the best overall results in terms of accuracy.
  • Compared to the best benchmarks, the model achieves a 0.17 point drop in RMSE, a 19.56% drop in MAPE, a 14.48% drop in SMAPE, and a 0.12 increase in Pearson’s correlation r. performance
    Source sentence
    The model achieves a 0.17 points drop in RMSE, 19.56% drop in MAPE, and 14.48% drop in SMAPE; it gains 0.12 in Pearson’s correlation r compared to the best performing benchmarks.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 8 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.