LSTM-Based Time Series Prediction for Bitcoin Price Analysis: ACase Study with Evaluation Metrics and Performance Insights

In this paper, we propose a time series forecasting model for predicting Bitcoin’s closingprices using a Long Short-Term Memory (LSTM) neural network. Bitcoin's priceprediction is a challenging task due to its volatile nature and the complex patterns in itshistorical data. The dataset used for training the model is the Bitcoin Historical Data,sourced from Kaggle, which includes minute-level trading data. To preprocess the data,MinMax scaling was applied to normalize the closing prices, and the dataset wasstructured into input-output pairs using a sliding window approach. The model architectureconsists of two LSTM layers followed by a Dense layer to output the predicted value. Themodel was trained with the Adam optimizer and Mean Squared Error (MSE) loss function.The performance of the model was evaluated using multiple metrics, including MeanAbsolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE),and R-squared (R²). The results indicate that the model achieved an MAE of X, MSE of Y,RMSE of Z, and an R² of A, demonstrating its ability to forecast Bitcoin’s closing priceswith reasonable accuracy. This work highlights the effectiveness of LSTM networks infinancial time series forecasting and lays the groundwork for future research incryptocurrency price prediction.

کلیدواژه ها:

Bitcoin priceprediction ، LSTM ، time seriesforecasting ، financial data ، cryptocurrency ، deeplearning.