The most well-known cryptocurrency, Bitcoin, has drawn a lot of interest because of its tremendous volatility and lucrative potential. Because of the impact of several variables, such as investor sentiment, market demand, regulatory changes, and macroeconomic conditions, it is difficult to predict Bitcoin values with any degree of accuracy. Deep learning algorithms are an appealing alternative for predicting since traditional financial models frequently fail to capture the intricate patterns in Bitcoin price fluctuations. This study explores the use of deep learning models, such as transformer-based models, convolutional neural networks (CNNs), and long short-term memory (LSTM) networks, to predict Bitcoin prices. Time-series forecasting is a good fit for LSTMs, a type of recurrent neural network (RNN), because of its exceptional performance at processing sequential data. While transformer-based architectures like attention mechanisms increase forecast accuracy by recognizing important elements in price patterns, CNNs, which are frequently employed in image processing, have demonstrated promise in catching local trends in financial data. Historical Bitcoin price data, including open, high, low, and closing (OHLC) values, trading volume, and other technical indicators, make up the dataset used in this study. To improve model performance, data preprocessing methods such feature engineering, outlier identification, and normalization are used. A training–validation–testing methodology is used in the study to assess how well the models forecast future Bitcoin prices. Model accuracy is tested using performance metrics like R2 score, mean absolute error (MAE), and root-mean-square error (RMSE). Various deep learning models are compared to identify the best successful strategy. Furthermore, the influence of outside variables on the prediction of the price of Bitcoin is investigated, including news sentiment and macroeconomic indices.