Cryptocurrency price prediction using GPR and SMOTE

Cryptography is used by cryptocurrencies to shift money without the intervention of centralized financial institutions. They are decentralized digital assets. On rapidly changing exchanges like those for crypto currencies, it is a tremendously taxing procedure for people to keep track of many simultaneous instantaneous price changes. As a solution to this, computer software that can make fast and objective decisions by constantly observing can replace humans. In this study, the closing price of Bitcoin (BTC), which has the highest volume in the crypto money system, is analyzed. In the study, in which the Gaussian Process Regression (GPR) model and the SMOTE method were used, data belonging to BTC for the period between 25/07/2010 and 05/06/2022 were used as the data set. Opening price, highest-lowest price, volume, dollar index and some indicators used in technical analysis were used as input parameters. The kfold method was followed in the separation of training and test data. The data is divided into 5 subsets with kfold. The mean MAPE value was found to be 1887, and the mean R2 value was found to be 0.99977 in the models with SMOTE. In addition, the GPR model and the GPR model functions that were applied to the SMOTE method were compared by excluding the opening price, which was the price that was highest-lowest, from the data. It was carried out to determine which model performed better.