Artificial Intelligence in Financial Forecasting: Comparing LSTM Networks and Traditional Econometric Models
1 viewsAbstract
In this project, I analyze the changing landscape of financial time series prediction by contrasting advanced deep learning models with more conventional statistical approaches. Historically, predictive models for finance have relied on linear methods like ARIMA and these models, by themselves, are insufficient to describe the nonlinear patterns of current financial timeseries, especially when the markets exhibit increasing volatility. I test the ability of LSTMs to compete with traditional statistical approaches, including econometrics-based forecasting and Random Forest models. I find that while deep learning models often require more data and computing power, the ability of deep learning models to accommodate multi-dimensional inputs increases forecast accuracy for many applications of finance
Keywords
ARIMA, Artificial Intelligence (AI), Econometrics, Financial Forecasting, LSTM Networks, Machine Learning