Recent advances in predictive models for stock market forecasting: a survey

Economic forecasting, especially stock market prediction, has received scholars’ attention because of the strong application background. With the emergence of new technologies in machine learning, deep learning, and statistical methods this year, methods for identifying trends in stock price movements have become more accurate. This paper aims at reviewing the current research on the current techniques used in prediction of stock price movements, whereby we consider both the conventional models including ARIMA as well as the modern machine learning algorithms which include the Random Forests, Support Vector Machines and the Gradient Boosting, the deep learning techniques such as LSTM and the CNN. Also, the incorporation of sentiment analysis and NLP approaches is investigated for improving the prediction performance using news articles and tweets as raw data. Comparing these models helps to identify their advantages and disadvantages, as well as appropriate fields of use, turning to future developments – boosting the real-time forecasts, making AI interpretations more understandable or integrating several data sources for more accurate predictions.