Fig 1.
Graphical overview of the proposed study.
Fig 2.
Brief Overview of the proposed Bitcoin prediction ACB-XDE framework.
Fig 3.
Proposed ACB-XDE detailed framework.
Fig 4.
Gating mechanism of LSTM.
Fig 5.
Component connection topology of LSTM.
Fig 6.
Traditional LSTM vs BiLSTM with an attention mechanism.
Fig 7.
Structure of new customized-attention mechanism.
Fig 8.
Stock price prediction methodology in the proposed ACB-XDE framework.
Fig 9.
Structure of the data.
Table 1.
Summary of data before and after preprocessing.
Table 2.
Outliers Z score greater than 3.
Fig 10.
Z-score analysis for outlier detection.
Fig 11.
Normalized features over time.
Table 3.
Attention-customized BiLSTM parameter settings.
Table 4.
Parameter settings of XGBoost.
Table 5.
Error Analysis of the proposed framework and evaluation with the state-of-the-art model.
Fig 12.
Bitcoin price prediction using BiLSTM with new attention.
Fig 13.
Bitcoin price prediction using XGBoost.
Fig 14.
The proposed ACB-XDE framework training and performance evaluation.
Fig 15.
Training and Validating the proposed ACB-XDE framework.
Fig 16.
Testing the proposed ACB-XDE framework.
Table 6.
Error analysis of the previous studies employed on bitcoin at various distributions.
Table 7.
MAE confidence interval variation.
Fig 17.
Comparison of the proposed ACB-XDE framework with state-of-the-art models.
Fig 18.
The sub-portion of the above Fig 17.
Fig 19.
The proposed ACB-XDE framework and state-of-the-art models gain.
Fig 20.
Difference of the proposed framework and state-of-the-art models with actual price per day.