An AI Robot-Analyst System for Investment Analysis

Abstract

Financial analysts have long been performing stock price forecasts. These analysts, being human, are limited by heuristics and biases which impact their forecast accuracy. Recent advances in machine learning research have opened the door to automation of human analysts' work and have rendered it faster and more efficient. In this study we propose an automated financial analyst that predicts stock prices one year ahead. We later compare its performance to human analysts. The robot-analyst utilizes a rich mixture of quantitative financial and stock related text data features fed into machine and deep learning algorithms to forecast one-year ahead stock prices. Comparing the robot-analyst’s predictions to the human ones revealed that, on average, the robot outperformed the human analysts by 19% in terms of RMSE. Simulation of investments done following the robot-analyst recommendations showed that our robot surpassed human analyst returns on most of the stocks we analyzed and led to higher rates of return.