Forecasting Financial Innovation Fund Prices : Machine Learning Approach

Recent innovations in financial technology (Fintech) have changed the landscape of financial services. Fintech is not limited to digital payment systems but also includes blockchain technology and cryptocurrency. Investors are increasingly looking for exposure to financial innovations. However, they are concerned about the high volatility, uncertainty, and risks embedded in the cryptocurrency market. Alternatively, investors can invest in equities related to financial innovation through exchanged traded funds (ETF). We focus on ETFs that are composed of firms associated with financial innovations and attempt to forecast future ETF prices using a machine learning process. Table 1 summarizes the ETFs examined in this study.We employ two machine learning algorithms and one standard econometric model for time-series forecasting: Long Short-Term Memory (LSTM), Prophet, and Auto-Regressive Integrated Moving Average (ARIMA). LSTM is advantageous for its capability to learn from long-term dependencies, whereas Prophet can handle seasonality and outliers effectively (Ning et al., 2022). The classical ARIMA model can address linear problems, revealing the dynamic force between the historical and predicted data (Zhao et al., 2022). Several recent studies have used these predictive models for time-series forecasting such as cash flow (Weytjens et al., 2021), Bitcoin (Rathore et al., 2022), and oil production (Ning et al., 2022). We apply these emerging machine learning models to financial innovation ETFs and evaluate the accuracy of the predictions using Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Square Mean Error (RSME). We contribute to the literature by applying machine learning-based predictive models to relatively new ETFs tied to financial innovation, in the hope that our results benefit retail investors and fund managers in constructing trading strategies. Our preliminary results suggest that Prophet is superior to other models