MACHINE LEARNING MODELING FOR PREDICTION OF BITCOIN MARKET PRICE USING A RECURRENT NEURAL NETWORK WITH LONG SHORT TERM MEMORY

Abstract

Machine learning based on neural networks has found applications in a wide range of sectors, including translation, finance, distribution, and the medical world, as well as cognition. This paper demonstrates a Recurrent Neural Network Learning Model based on LSTM that evaluates previous Bitcoin prices and predicts the upcoming one. Statistical analysis/methods: This model predicts the actual and anticipated Bitcoin prices for the following 81 days by learning the previous 30 days' prices and then forecasting the next day's price. The regularised data set for modelling is separated into test and training data sets at a 1:9 ratio. The latter set is divided once more into training and verification data. This study's Machine Learning will require the use of a Neural Network library and the Keras framework. Findings: Fitting the model entails determining the model's weight by optimising the procedure while using training data. In this paper, the batch size of the fit function is 11 and the number of epochs is 30. As learning is processed more frequently, the loss declines more monotonously, and it eventually converges to a more regular value. In other words, there is no overfitting.