Hybrid metaheuristic algorithm methods and econometric models in prediction of dogecoin price

Purpose - This paper aims to predict Dogecoin price by using artificial intelligence (AI) methods and comparing the results with the econometrics models. Design/methodology/approach - An artificial neural network (ANN) was applied as a prediction method without any optimization techniques. Additionally, the genetic algorithm (GA) is used to select the most appropriate input variables. Additionally, based on the literature review and the relationships between cryptoprice and global indices, 20 economic indicators, such as Coinbase Bitcoin, Coinbase Litecoin and US dollars, along with main global stock indices such as FTSE100 and NIFTY50, are identified as input variables for the model. Lichtenberg algorithm (LA) and aquila optimization (AO) algorithm are used to make the ANN more robust. To validate our algorithms, they have been implemented on daily data for the last three years. To demonstrate the superiority of the models over traditional methods such as econometrics, regression analysis and curve fitting techniques are used. The effectiveness of these models is then evaluated and compared using criteria such as recall, accuracy and precision. Findings - The results indicate that AI-based algorithms not only enhance the accuracy, recall and precision of calculations but also expedite the process without requiring the numerous and restrictive assumptions associated with time series and econometric models. Originality/value - The main contribution of this paper is the application of novel approaches such as AO and LA to improve the predictive capabilities of the ANN method for various cryptocurrencies' prices. It demonstrates the superiority of the proposed algorithms over traditional econometric models using real-life data.