<p>This study presents a deep learning paradigm for Ripple (XRP) price forecasting with technical indicators, dynamic feature weighting, and hybrid neural networks as its key constituents. Many technical indicators, including MACD, RSI, OBV, Bollinger Bands, etc., are evaluated to capture the historical price signals and market behavior. For improved predictive performance, two hybrid architectures are proposed, CNN-LSTM-GRU with Attention and CNN-GRU-LSTM with Attention, in which the CNN acts as a local feature extractor, the LSTM and GRU perform sequential learning, and the attention mechanisms highlight important time steps. The models are tested under various Pearson’s thresholds of technical indicators, and performance is measured on RMSE, MAPE, and R2. The results of the experiments indicate that the CNN-GRU-LSTM-Attention model is consistently superior in terms of lower error rates and higher accuracy compared to the independent models when measuring the same forecast. Dynamic input feature weighting facilitates the model to learn both spatial and temporal interactions underlying volatile markets. The indicators do not always provide consistent results across thresholds, but the proposed ensemble method shows a strong generalization. Future improvements will include optimizing the feature space with evolutionary algorithms and the addition of macroeconomic impact for better long-term forecasts. This research contributes to AI-based cryptocurrency prediction by developing hybrid models while integrating technical indicators.</p>