Comprehensive Analysis of Machine Learning Algorithms Used in Robo-Advisory Services

Abstract

Adoption of Robo advisory is rapidly increasing among retail investors. Approximately 10% of all assets under management will be managed by them by 2025. Researchers continuously work on finding machine learning and deep learning algorithms which will accurately predict the stock price movement with less error. Various Robo advisor platform providers are working on different approaches like NLP (Natural language algorithm) based sentiment analysis, hybrid time series algorithms to limit human bias, which existed in the current financial advisory services. This research compared the performance of various forecasting algorithms, which can be used in the Robo advisory framework. RMSE (Root mean square value) value is taken into consideration in this research to compare the performance of algorithms as it is a good and simple loss function. It has been observed that newer algorithms like those proposed by Facebook are able to predict stock prices pretty well, but ARIMA outperformed all the algorithms which are considered in this research.