Deep Learning Algorithms for MIR4 – Draco Token Exchange Value Forecasting 

MIR4, is a play to earn game that uses Non-Fungible Tokens (NFT) and cryptocurrency—or in MIR4, Draco Tokens- as a reward. Draco is obtained through mining an in-game resource called Dark Steel and is then traded to Wemix Wallet, where real-world money is obtained. Cryptocurrencies are volatile, which gives MIR4 players and traders a decision dilemma of when is the preferable time to buy, sell, or trade Draco tokens. In this study, we present Deep Learning models specifically the Long – Short Term Memory (LSTM) neural network, and NeuralProphet (NP) time series machine learning models to forecast a 10-day ahead Draco token exchange value. Historical data of Draco value is utilized as a univariate input for the analysis, model development, and forecasting of the future values of Draco token exchange value. Performance of formulated models are assessed and compared based on the following regression metrics: RMSE, MSE, MAE and MAPE. Experimental results indicate that the LSTM Neural Network yielded better forecast estimates with lower error than the NeuralProphet with evaluation metric scores of RMSE = 0.009, MSE = 0.010, MAPE = 0.224 and MAE = 0.077. Findings of this study showed that LSTM can be utilized as a tool for forecasting future Draco token exchanged values. Heuristic, hybrid and other machine learning models may be explored for an accurate forecasting.

Keywords: mir4, draco token, deep learning, LSTM, NeuralProphet, forecasting

Corresponding author's email: gomez.neilarchein21@gmail.com