Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Expert System for Bitcoin Forecasting: Integrating Global Liquidity via TimeXer Transformers

Sravan Karthick T

TL;DR

Bitcoin price forecasting faces extreme volatility and non-stationarity, limiting traditional univariate models for longer horizons. The paper introduces TimeXer, a Transformer-based architecture that incorporates exogenous macro drivers by embedding Global M2 Liquidity with a dense lag structure and cross-attention to the endogenous Bitcoin series. Across 2020–2025 data and a 70-day forecast horizon, TimeXer-Exog delivers substantial gains (e.g., $1.08e8$ MSE at 70 days, ~89% better than the univariate baseline) and shows robustness via the Model Confidence Set, indicating that macro-financial conditioning yields more stable, directionally coherent forecasts. The findings imply practical benefits for risk management and algorithmic trading in crypto markets and motivate future work with additional macro variables and cross-asset validation.

Abstract

Bitcoin price forecasting is characterized by extreme volatility and non-stationarity, often defying traditional univariate time-series models over long horizons. This paper addresses a critical gap by integrating Global M2 Liquidity, aggregated from 18 major economies, as a leading exogenous variable with a 12-week lag structure. Using the TimeXer architecture, we compare a liquidity-conditioned forecasting model (TimeXer-Exog) against state-of-the-art benchmarks including LSTM, N-BEATS, PatchTST, and a standard univariate TimeXer. Experiments conducted on daily Bitcoin price data from January 2020 to August 2025 demonstrate that explicit macroeconomic conditioning significantly stabilizes long-horizon forecasts. At a 70-day forecast horizon, the proposed TimeXer-Exog model achieves a mean squared error (MSE) 1.08e8, outperforming the univariate TimeXer baseline by over 89 percent. These results highlight that conditioning deep learning models on global liquidity provides substantial improvements in long-horizon Bitcoin price forecasting.

Expert System for Bitcoin Forecasting: Integrating Global Liquidity via TimeXer Transformers

TL;DR

Bitcoin price forecasting faces extreme volatility and non-stationarity, limiting traditional univariate models for longer horizons. The paper introduces TimeXer, a Transformer-based architecture that incorporates exogenous macro drivers by embedding Global M2 Liquidity with a dense lag structure and cross-attention to the endogenous Bitcoin series. Across 2020–2025 data and a 70-day forecast horizon, TimeXer-Exog delivers substantial gains (e.g., MSE at 70 days, ~89% better than the univariate baseline) and shows robustness via the Model Confidence Set, indicating that macro-financial conditioning yields more stable, directionally coherent forecasts. The findings imply practical benefits for risk management and algorithmic trading in crypto markets and motivate future work with additional macro variables and cross-asset validation.

Abstract

Bitcoin price forecasting is characterized by extreme volatility and non-stationarity, often defying traditional univariate time-series models over long horizons. This paper addresses a critical gap by integrating Global M2 Liquidity, aggregated from 18 major economies, as a leading exogenous variable with a 12-week lag structure. Using the TimeXer architecture, we compare a liquidity-conditioned forecasting model (TimeXer-Exog) against state-of-the-art benchmarks including LSTM, N-BEATS, PatchTST, and a standard univariate TimeXer. Experiments conducted on daily Bitcoin price data from January 2020 to August 2025 demonstrate that explicit macroeconomic conditioning significantly stabilizes long-horizon forecasts. At a 70-day forecast horizon, the proposed TimeXer-Exog model achieves a mean squared error (MSE) 1.08e8, outperforming the univariate TimeXer baseline by over 89 percent. These results highlight that conditioning deep learning models on global liquidity provides substantial improvements in long-horizon Bitcoin price forecasting.
Paper Structure (24 sections, 4 equations, 5 figures, 2 tables)

This paper contains 24 sections, 4 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Bitcoin forecasting using ARIMA
  4. Bitcoin forecasting using CNN, RNN, and LSTM models
  5. Bitcoin forecasting using Transformer-based models
  6. Incorporating exogenous variables in Bitcoin forecasting
  7. Methodology
  8. The TimeXer Architecture
  9. Exogenous Input Integration and Cross-Attention.
  10. Baseline Models
  11. Bitcoin and Global Liquidity Relationship
  12. Statistical Evaluation: Model Confidence Set (MCS)
  13. Data and Experimental Setup
  14. Data Description
  15. Experimental reference date and temporal alignment.
  16. ...and 9 more sections

Figures (5)

  • Figure 1: The TimeXer architecture showcasing the integration of endogenous and exogenous data streams through specialized embedding and attention mechanisms.
  • Figure 2: Bitcoin open price v/s Global Liquidity with a 12-week lead (both are standard normalized)
  • Figure 3: Forecast comparison at horizon $H=7$.
  • Figure 4: Forecast comparison at horizon $H=28$.
  • Figure 5: Forecast comparison at horizon $H=63$.