Real-world model for bitcoin price prediction. Issue 4 (July 2022)

• Record Type: Journal Article
• Authors: Rathore, Rajat Kumar; Mishra, Deepti; Mehra, Pawan Singh; Pal, Om; HASHIM, AHMAD SOBRI; Shapi'i, Azrulhizam; Ciano, T.; Shutaywi, Meshal
• Is Part Of: Information processing & management. Volume 59:Issue 4(2022)
• Publication Date: 2022-07
• Subjects: Bitcoin -- Cryptocurrency -- Machine learning -- Prediction -- Time series analysis -- Fbprophet model; Information storage and retrieval systems -- Periodicals; Information science -- Periodicals; Systèmes d'information -- Périodiques; Sciences de l'information -- Périodiques; Information science; Information storage and retrieval systems; Periodicals; 658.4038
• DOI: 10.1016/j.ipm.2022.102968
• Languages: English
• ISSNs: 0306-4573

Abstract

Highlights: The aim is to achieve a model to predict closing price of bitcoin along with bitcoin opening Price, bitcoin day high Price, bitcoin day low Price, bitcoin day volume and market capitalization of bitcoin on day by using deep leaning algorithms and various concepts of machine learning, which can find hidden patterns in data, combine them, and make more accurate predictions. Majorly two machine learning algorithms are proposed and implemented for forecasting bitcoin values. Built a time-series model for which fbprophet library is being used. The fbprophet model is utlised to predict real-world outcomes after eliminating the seasonality effect. Abstract: Cryptocurrency is a new sort of digital asset that has evolved as a result of advances in financial technology, and it has provided a significant research opportunity. There are many algorithms for price prediction for crypto currencies like LSTM and ARIMA. However, the downside is that LSTM-based RNNs are difficult to comprehend, and gaining intuition into their behavior is tough. In order to produce decent outcomes, rigorous hyperparameter adjustment is also essential. Furthermore, crypto currencies do not precisely adhere to past data, and patterns change fast, reducing the accuracy of predictions. Cryptocurrency price forecasting is difficult due to price volatility and dynamism. Because the data is dynamic and heavily influenced by various seasons, the ARIMA model is unable to handle seasonal data. In order to provide better price predictions for crypto traders, a new model is required. The objective of the study is to apply Fbprophet model as the key model because it is superior in functionality as compared to LSTM and ARIMA additionally removing the pitfalls generated in LSTM and ARIMA model while analyzing the cryptocurrency data. This study provides a methodology for predicting the future price of bitcoin that does not rely solely on past data due to seasonality in historical data. So, after fitting the seasonality and smoothing, the model is constructed that can be useful for real-world use cases. In case of crypto currencies where less historical data is available and it is hard to find pattern, proposed method can easily deal this type of problems. Overall difference between predicted and actual values is low as compared to other model even after seasonal data was available.