Comparative Analysis of LSTM and GRU Models with Hyperparameter Optimization for Bitcoin Price Prediction

Forecasting Economically Significant Bitcoin Moves: A Multi-Scale TCN with Profit-Optimized Thresholds

Bitcoin's future fluctuations are a substantial concern for investments and risk management. Investors and financial institutions require accurate forecasts of these price movements to hedge and optimize portfolios. This study aims to answer the question of whether Bitcoin's price will rise beyond 5% within the next 7 days by utilizing on-chain, market, and sentiment data from February 2018 to December 2025. The proposed model consists of a multi-scale temporal convolutional network with InceptionTCN blocks, CNN channel attention, adaptive average pooling, and a pairwise ranking loss. Dilated convolutions with bottleneck and fusion layers are employed to efficiently capture features over horizons from 1 to 4 days. Given the class imbalance in the dataset, AUC is used instead of accuracy and other classification metrics to reflect the model's performance better. Subsequently, a profit-optimized decision threshold is also applied to align model selection with financial objectives. The proposed model is compared with 5 other baselines: ImprovedTCN_GRU, LSTM, TCN, XGBoost, and Random Forest. Results indicate that the proposed model achieved an AUC of 0.6316 and a profit of 1.703, outperforming all baseline models. Using a novel deep learning model would assist investors in making better financial decisions.

A hybrid FinBERT-LSTM framework for Bitcoin price forecasting using news sentiment and technical indicators

This study proposes a hybrid forecasting framework that integrates sentiment analysis with deep learning to predict Bitcoin’s hourly and daily closing prices. Hourly BTC/USD market data spanning June 2021 to November 2025 were combined with approximately 326,000 Bitcoin-related news headlines published over the same period. Sentiment scores in the range of [-1, +1] were generated for each headline using FinBERT, a transformer-based language model trained on financial texts, and were subsequently integrated with technical indicators such as trading volume, MACD, and RSI. The resulting combined feature set was modeled using an LSTM network to capture temporal dependencies. Empirical results demonstrate that sentiment-enhanced hybrid models consistently outperform models based solely on technical indicators across RMSE, MAE, MAPE, and R² metrics. The hourly hybrid model achieved the best performance, with an RMSE of 1,009 USD and an R² of 99.23%. Furthermore, a 30-day out-of-sample real-time evaluation yielded an RMSE of 941 USD. The consistency between in-sample and out-of-sample results indicates that the proposed framework maintains stable predictive performance over time.

FORECASTING BITCOIN PRICE RANGES BY COMBINING SENTIMENT ANALYSIS WITH HIGH-DIMENSIONAL INDICATORS IN AN LSTM FRAMEWORK

In this paper, I suggest a way to estimate the range of Bitcoin prices for the following day by combining Long Short-Term Memory (LSTM) networks with natural language processing (NLP) approaches. To capture the complex sentiment of the market, the model combines sentiment elements that are collected from Twitter data using sophisticated natural language processing techniques with high-dimensional technical indicators. The LSTM model successfully learns temporal correlations and intricate patterns by integrating sequential analysis of both numerical market indicators and textual sentiment data, improving its capacity to predict changes in Bitcoin prices. The trials make use of millions of pertinent Twitter messages and six years' worth of Bitcoin market data. The method shows how sentiment analysis and deep learning architectures can be combined to increase forecasting resilience and interpretability in erratic cryptocurrency markets. Sensitivity analysis is used to maximise the impact of sentiment characteristics, emphasising the value of sentiment-driven insights in financial prediction models and providing a fresh viewpoint for more precise and dynamic forecasting of the cryptocurrency market.

Comparison of neural network models for prediction cryptocurrency price volatility in trading pairs

Crypto Price Analysis: An AI Perspective

Accurate prediction of Bitcoin market prices is challenging because of extreme volatility, nonlinear dynamics and rapid sentiment-driven fluctuations observed in the cryptocurrency market, especially Bitcoin. In this research, the authors investigate the use of machine learning and deep learning methods for forecasting Bitcoin prices using historical market data, technical indicators, and sentiment analysis from social media. A comprehensive dataset was used to develop and test three experimental setups from 2015 to 2024. The first configuration, an LSTM model was trained on historical OHLCV (Open, High, Low, Close, Volume) data only. In the second setup, the model was improved with the inclusion of technical indicators based on market behavior, including momentum indicators and trend indicators. The third configuration used social media sentiment features (extracted from over 2.5 million tweets related to Bitcoin, employing polarity and subjectivity scoring via TextBlob), added to the previous configuration. The results show that the LSTM model with technical indicators gave the best prediction performance with R2 of 0.9084 and Mean Absolute Error (MAE) of 0.0387, which is 8.5% better than the historical data-only model. By comparison, the accuracy of prediction did not significantly improve with the addition of Twitter sentiment features. The feature importance analysis also revealed that Relative Strength Index (RSI) and MACD histogram were the most significant features in predicting the price movements of Bitcoin. The study brings three significant findings: (1) a methodological framework based on LSTM neural networks for comparing the performance of various sets of features in predicting Bitcoin prices; (2) the discovery of the most relevant features in cryptocurrency markets; and (3) proof that technical indicators have a significant impact on the predictive accuracy of this model, while social media sentiment has a minor impact under the circumstances. The results offer valuable implications for researchers and practitioners building data-driven models for predicting cryptocurrencies.

LSTM NETWORK FOR HIGH-VOLATILITY STOCK PRICE PREDICTION: TESLA INC

This research investigates the efficacy of Long Short-Term Memory (LSTM) networks for predicting stock prices of high-volatility equities, with application to Tesla Inc. (TSLA). Addressing a gap in financial machine learning literature, we develop an advanced LSTM architecture trained on Tesla's daily closing prices from January 1, 2017, to November 20, 2024. Through meticulous preprocessing, strategic dropout regularization, and sophisticated sequence modeling, our model achieves a Root Mean Square Error (RMSE) of $12.17 and a Mean Absolute Error (MAE) of $8.51. For comparative purposes, we implement a walk‑forward ARIMA benchmark, which achieved an RMSE of $8.30 and an MAE of $5.76, indicating superior point forecast accuracy. However, the LSTM model demonstrated better directional accuracy (50.60% against 47.62%), suggesting complementary strengths across evaluation metrics. The Diebold‑Mariano test confirmed a statistically significant difference between the two models (DM = 4.58, p < 0.01). This study contributes to the understanding of deep learning applications in financial markets and establishes new benchmarks for volatile stock prediction. The findings support adopting hybrid approaches that combine the point forecast accuracy of traditional econometric models with the directional predictive capabilities of deep learning architectures for financial forecasting in turbulent market conditions.