In this section we represent the simulation result for the prediction of the prices of the two cryptocurrencies for next 10 days. The 2 coins that we have used for the current studies are: - XRP and Chainlink. XRP contains 2893 data points and Chainlink contains 1385 data points. The sample dataset used for the prediction is different than the data set used to train the model for reason to judge the actual performance of the designed model.

For each of the coin i.e., XRP and Chainlink different metrics such as MSE (Mean Square Error), RMSE (Root Mean Square Errors), MAE (Mean Absolute Error), Regression Score, R2 Score, MGD (Mean Gama Deviance Regression Loss) and MPD (Mean Poisson Deviance Regression Loss) was taken on note and compared to find out the optimal model for prediction of XRP and Chainlink.

The Following Table-5,6 shows comparison of evaluation matrix results prediction by Stacked Lstm, GRU and our proposed approach (LSTM + GRU, GRU + LSTM).

Table 5: Training Data Results of LSTM, GRU and Proposed Approach

Table 6: Testing Data Results of LSTM, GRU and Proposed Approach

The RMSE (Root Mean Square Error) for XRP using GRU + LSTM hybrid Model came as 0.0535 followed by, LSTM + GRU hybrid model (0.0569), Stacked LSTM (0.0599) and GRU (0.0658). In the Chainlink on application of LSTM + GRU hybrid model produced a RMSE of 0.5697 followed by GRU (5.2578), GRU + LSTM (6.8070) and Stacked LSTM (7.2339)

Root Mean Square Error (RMSE) provides for a measure between the Predicted price and actual price of the coins. Lower values of RMSE signifies better fit of the model for the prediction. From the above data we can infer that the hybrid model GRU + LSTM performs better than other models while predicting the price of the coins XRP and LSTM + GRU hybrid model performs better than other models while predicting the price of Chainlink.

The MSE (Mean Square Error) for XRP using GRU + LSTM hybrid Model came as 0.0028 followed by, LSTM + GRU hybrid model (0.0032), Stacked LSTM (0.0035) and GRU (0.0043). Accordingly, in the Chainlink on application of LSTM + GRU hybrid model produced a MSE of 0.0032 followed by GRU (27.6445), GRU + LSTM (46.3358) and Stacked LSTM (52.3293).

Similar result is obtained while comparing the MAE values for the coins using different models. For XRP, GRU + LSTM shows less MAE value (0.0227), followed by LSTM + GRU (0.0294), Stacked LSTM (0.0329) and GRU (0.0347). For Chainlink MAE value of 0.0294 produced by LSTM + GRU hybrid model followed by GRU (3.8989), GRU + LSTM (5.1457) and Stacked LSTM (5.4561).

In MAE, the absolute difference between the Predicted price and the actual price of the coins are taken in account and then the average of these absolute differences is compute. Although it is less sensitive to outliers compared to RMSE, as it does not square the difference, but can be used marginally to infer the inconsistency, if any, in RMSE calculations. From our results, RMSE and MAE relates to each other and produces the same analysis, thereby terming GRU + LSTM is the best fit model to predict the price of XRP coins and LSTM + GRU is the best fit model to predict the price of Chainlink.

Furthermore, by analyzing the R2 Score, the GRU + LSTM hybrid model possesses the highest score of 0.9696, closely followed by the LSTM + GRU (0.9656), Stacked LSTM (0.9619), and GRU (0.9540), models suggesting a good model fit for XRP coin. Similarly, the LSTM + GRU hybrid model has the highest R2 Score of 0.9656, followed closely by the GRU model with an R2 score of 0.9635, indicating these models fit the data better for Chainlink predictions.

The analysis above is further confirmed by analyzing MGD (Mean Gamma Deviance Regression Loss) and MPD (Mean Poisson Deviance Regression Loss. The hybrid model, GRU + LSTM consistently stands out across MGD, MPD and Regression Score, reinforcing its robustness in prediction in case of XRP coin and LSTM + GRU hybrid model consistently stands out across MGD, MPD and Regression Score, reinforcing its robustness in prediction in case of Chainlink coins.

GRU + LSTM and LSTM + GRU both outperform in terms of RMSE and MAE showing a marginal advantage in R2 score in case of XRP Coin.

The LSTM + GRU model provides the best performance across all metrics. However, the competition remains tight, especially between GRU and GRU + LSTM in RMSE and MAE metrics.

When comparing individual models against combined architectures, it becomes evident that the hybrid structures, GRU + LSTM, LSTM + GRU tend to showcase superior predictive prowess in XRP and Chainlink respectively.

Both the hybrid models continue to be the front-runners in different case. Their consistently top-tier performance across most metrics is indicative of the potential advantages of combining the attributes of both LSTM and GRU.

The comparison graphs of original close price with predicted close price using 15a,15b,15c,15d,15e,15f,15g,15h.

After getting the comparison graph the predicted close price graph of Stacked LSTM, GRU with two proposed model has been shown in Figure- for next 10 days has been shown in Figure-16a,16b,16c,16d,16e,16f,16g,16h