Paper: arXiv 2501.13136

Authors: Ramin Mousa, Meysam Afrookhteh, Hooman Khaloo, Amir Ali Bengari, Gholamreza Heidary

Abstract

Digital currencies have become popular in the last decade due to their non-dependency and decentralized nature. The price of these currencies has seen a lot of fluctuations at times, which has increased the need for prediction. As their most popular, Bitcoin(BTC) has become a research hotspot. The main challenge and trend of digital currencies, especially BTC, is price fluctuations, which require studying the basic price prediction model. This research presents a classification and regression model based on stack deep learning that uses a wavelet to remove noise to predict movements and prices of BTC at different time intervals. The proposed model based on the stacking technique uses models based on deep learning, especially neural networks and transformers, for one, seven, thirty and ninety-day forecasting. Three feature selection models, Chi2, RFE and Embedded, were also applied to the data in the pre-processing stage. The classification model achieved 63% accuracy for predicting the next day and 64%, 67% and 82% for predicting the seventh, thirty and ninety days, respectively. For daily price forecasting, the percentage error was reduced to 0.58, while the error ranged from 2.72% to 2.85% for seven- to ninety-day horizons. These results show that the proposed model performed better than other models in the literature.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced deep learning techniques (transformers, stacked models, wavelet denoising) but is heavily focused on data processing and real-world forecasting results with specific metrics, making it both mathematically complex and empirically rigorous.

Research Flowchart

  flowchart TD
  A["Research Goal: Predict Bitcoin<br>Prices & Movements"] --> B["Preprocessing<br>Wavelet Denoising & Feature Selection"]
  B --> C["Data: BTC Price &<br>Hashrate Features"]
  C --> D["Stack Deep Learning Model<br>Deep Learning + Transformers"]
  D --> E{"Forecast Horizons"}
  E --> F["1, 7, 30, 90 Days"]
  F --> G["Classification Accuracy:<br>63% to 82%"]
  F --> H["Regression Error:<br>0.58% to 2.85%"]

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 stacked deep learning classification and regression model that uses wavelet denoising to predict Bitcoin price movements and prices at multiple horizons.

Source sentence
This research presents a classification and regression model based on stack deep learning that uses a wavelet to remove noise to predict movements and prices of BTC at different time intervals.

Assumes 6

  • Training data consists of 80% of the total dataset and test data 20%. data
    Source sentence
    Specifies the training dataset(80% of the total data).
  • Test data consists of the remaining 20% of the total dataset. data
    Source sentence
    Test data: This part of data is used to evaluate the built model(20% of the total data).
  • Missing values are imputed using linear interpolation. data
    Source sentence
    In preprocessing, the linear interpolation method imputes missing value items as much as possible.
  • Outliers are controlled using the Isolation Forest algorithm. data
    Source sentence
    On the other hand, the Isolation Forest method algorithm [36] was used to control outliers.
Show 2 more
  • Binary price direction labels: 1 for increase, 0 for decrease. model
    Source sentence
    price of the next step was set to 1 in case of increase and 0 in case of decrease.
  • Wavelet denoising effectively removes noise from the price series. model
    Source sentence
    uses a wavelet to remove noise to predict movements and prices of BTC

Data

Backtest sample: BTC/USDT · crypto · daily

  • Bitcoin Dataset crypto-onchain public
    [2013/04/01 - 2019/12/31] (three interva · BTC hash rate and on‑chain features
    Source sentence
    Table 2 shows the range of each of these intervals.

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The classification model achieved 63% accuracy for next‑day prediction and 64%, 67% and 82% for 7‑, 30‑ and 90‑day horizons respectively. predictive-accuracy
    Source sentence
    The classification model achieved 63% accuracy for predicting the next day and 64%, 67% and 82% for predicting the seventh, thirty and ninety days, respectively.
  • Daily price forecasting error was reduced to 0.58% MAPE, with 2.72%–2.85% MAPE for 7‑ to 90‑day horizons. performance
    Source sentence
    For daily price forecasting, the percentage error was reduced to 0.58, while the error ranged from 2.72% to 2.85% for seven- to ninety-day horizons.
  • Using Chi2 feature selection, classification accuracy reached 62±1% for interval I, 60±1% for interval II and 57±1% for interval III. predictive-accuracy
    Source sentence
    In interval I, the Chi2 feature selection approach reached an accuracy of 62±1
  • For regression, Chi2 feature selection yielded MAE 1.21±0.01, RMSE 1.85±0.02, MAPE 0.50±0.02 in interval I. performance
    Source sentence
    In interval I, the Chi2 feature selection approach reached MAE = 1.21±0.01, RMSE = 1.85±0.02, and MAPE = 0.50±0.02.
  • The deep stack model outperformed baseline models (ANN, SVM, LSTM, etc.) in MAE, RMSE and MAPE across all intervals. performance
    Source sentence
    Fig. 7: MAE, RMSE, and MAPE of the price prediction of the proposed models and baseline for forecast in Interval I, II, and III.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 51 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.