Stock indices price prediction in real time data stream using deep learning with extra-tree ensemble optimisation

Stock prices follow random walk motion and are highly volatile. Earlier prediction models that use 'machine learning (ML)' and neuro-computational techniques for forecasting stock prices are more complex and less accurate. In this work, a novel deep learning with extratree ensemble (DELETE) optimisation for predicting stock indices price trends in real-time data stream is proposed. Each decision tree in the extra-tree (ET) forest selects k best feature to optimise the loss. ET ensemble aggregates the decisions from multiple de-correlated decision trees, thus normalising the total reduction in optimisation parameter. Finally, k highly predictive stock technical indicators (STIs) have been selected to supply as tensor to model. The model performance has been evaluated over three benchmark classifiers with three popular National Stock Exchange (NSE) indices. The daily prediction model achieved an accuracy up to 78.9% and average accuracy of 66.61%, which is up to 30.2% higher than benchmark models.