MVLSC Home · Issue Contents · Forthcoming Papers
Forecasting Bitcoin Using Monte Carlo Methods and Feedforward Neural Networks
Perla Motola-Villanueva and Klender Cortez
Cryptocurrencies have emerged as a disruptive investment, offering opportunities for investors. This study assesses the financial viability of investing in Bitcoin by analyzing potential price fluctuations over a five-year horizon. The research employs Monte Carlo simulations based on four stochastic models: Geometric Brownian Motion, Jump Diffusion Model, Heston Model, and Feedforward Neural Network.
A core component of the analysis involves a back test using historical data from 2020 to 2025. Each model is applied to analyze real price movements to determine which one most accurately captures the behavior and volatility of Bitcoin, incorporating key events such as Bitcoin halving. The results indicate that the Feedforward Neural Network exhibited superior predictive power compared to the traditional stochastic methodologies. As a result, this study contributes to assessing which projection methodology provides a more accurate simulation of Bitcoin prices relative to traditional forecasting models used in traditional financial markets.
Keywords: Bitcoin, blockchain, Feedforward Neural Network, halving, machine learning, Monte Carlo simulations, stochastic models