Abstract
In recent years, digital currencies, also known as cybercash, digital money, and electronic money, have gained significant attention from researchers and investors alike. Cryptocurrency has emerged as a result of advancements in financial technology and has presented a unique opening for research in the field. However, predicting the prices of cryptocurrencies is a challenging task due to their dynamic and volatile nature. This study aims to address this challenge by introducing a new prediction model called Bayesian optimization with stacked sparse autoencoder-based cryptocurrency price prediction (BOSSAE-CPP). The main objective of this model is to effectively predict the prices of cryptocurrencies. To achieve this goal, the BOSSAE-CPP model employs a stacked sparse autoencoder (SSAE) for the prediction process and resulting in improved predictive outcomes. The results were compared to other models, and it was found that the BOSSAE-CPP model performed significantly better.
TopIntroduction
Cryptocurrency, as defined by Aljadani (2022), is a digital or virtual form of currency that is utilized in financial systems. It is safeguarded by cryptography, making it difficult to be duplicated or counterfeited. Unlike traditional currency, it is not controlled by central banks or a central authority, and is instead distributed through cryptographic procedures (Ammer & Aldhyani, 2022). This distinguishes it from conventional currency. One of the unique features of cryptocurrency is that it is generated through a technique called blockchain, which is highly complex and designed to safeguard data by storing it in a manner that is challenging to hack, alter, or defraud the network (Anandkumar, 2021). Bitcoin, the first and most well-known cryptocurrency, has begun to establish a position for itself that could either aid in the widespread acceptance of cryptocurrency or be the cause of its downfall (Bačanin Džakula, 2021). Although cryptocurrency is still in its infancy, it is unclear whether or not it will find widespread acceptance on the worldwide market. Currently, there are approximately 5.8 million active users and 5,000 different cryptocurrencies in the cryptocurrency field (Cho et al., 2021). Due to its combination of encryption techniques with financial units, Bitcoin has recently gained interest in the fields of computer science, economics, and cryptography.
The use of cryptocurrency, specifically Bitcoin, has gained significant attention in recent years due to its unique characteristics of utilizing encryption techniques in financial systems (Daulton et al., 2022). Unlike traditional currency, cryptocurrency is not controlled by central banks or authorities and is instead distributed through complex cryptographic procedures, making it resistant to counterfeiting and double-spending (Daulton et al., 2022). Additionally, the technology behind cryptocurrency, known as blockchain, aims to make information storage secure and difficult to hack or alter (Daulton et al., 2022).
Despite the growing popularity of Bitcoin and other cryptocurrencies, there are still significant risks associated with their sustainability (Gupta & Nalavade, 2022). The value of Bitcoin is often affected by transaction fees, with companies and individuals encouraging the purchase of Bitcoin by offering lower transaction rates. However, this can also lead to an increase in transaction fees, resulting in a decrease in the benefits of using Bitcoin as compared to traditional fiat currency (Hitam et al., 2022). Recent advancements in deep learning (DL) have the potential to address these issues, as DL and neural networks (NN) have demonstrated success in a variety of fields such as education, pattern recognition, industry, and healthcare (Luo et al., 2022). DL techniques, in particular, have shown to be more efficient than traditional machine learning (ML) methods, and are able to handle complex and disordered data with various labeling errors and variables (Maleki et al., 2020).
Previous studies have proposed using DL-based techniques such as Bi-LSTM and GRU for forecasting the prices of common cryptocurrencies, such as Cardano, Bitcoin, and Ethereum (Mohanty & Dash, 2022). The use of the Grid Search system for hyperparameter optimization and meta-heuristic techniques such as the chaotic Dolphin Swarm Optimized technique have also been suggested to improve forecast accuracy (Petrovic et al., 2021). Another study proposed using ML techniques to determine the optimal method for predicting Bitcoin prices based on the prices of other popular cryptocurrencies, such as Zcash, Litecoin, and Ethereum (Salb et al., 2022).
Key Terms in this Chapter
Price Prediction: Price prediction refers to the forecasting or estimation of the future price of a particular asset, commodity, or currency, based on various factors such as market trends, demand, and supply.
Stacked Sparse Autoencoder (SSAE): Stacked Sparse Autoencoder (SSAE) is a deep learning architecture that consists of multiple autoencoder layers, each layer being trained to learn representations of the input data in a compact and sparse format.
Operates Independently: It refers to an action or system that functions without outside influence or control.
Machine Learning (ML): Machine learning is a subfield of artificial intelligence that focuses on the development of algorithms and models that enable computers to learn and make predictions or decisions based on data.
ARIMA, LSTM, and GRU Models: ARIMA stands for Autoregressive Integrated Moving Average, LSTM stands for Long Short-Term Memory, and GRU stands for Gated Recurrent Unit. ARIMA, LSTM, and GRU models are popular time series prediction algorithms commonly used in the field of machine learning for forecasting future values based on historical data.
AE Learning Neural Network: AE (Autoencoder) learning neural network refers to a type of machine learning model that aims to recreate the input data in an unsupervised manner through a series of hidden layers. The objective of an AE learning neural network is to learn the underlying patterns in the data by reconstructing the input data with high accuracy.
Cryptocurrency: Cryptocurrency is a digital or virtual currency that uses cryptography for security and operates independently.
Hyperparameter Values: Hyperparameter values refer to the adjustable settings in a machine learning model that are set before training the model and have a direct impact on its performance.
Reconstructed Error (RE): The Reconstructed Error (RE) is a measure of the difference between the actual value and the predicted value in a machine learning model, often used as a metric for evaluating the performance of autoencoder models.
DNN Models: DNN stands for Deep Neural Network models, which are a type of artificial neural network with multiple hidden layers used for complex data analysis and predictions.
Bayesian Optimization (BO): Bayesian Optimization (BO) is a probabilistic model-based approach to optimization problems, where the objective function is unknown but can be evaluated at any given point.