We have built an enterprise-grade AI engine that specialises in large-scale financial time-series forecasting on multiple instruments. Our proprietary technology is based on our skill set in combining the fields of artificial intelligence and financial engineering. Here are a few of the key technologies that we use to achieve higher forecasting accuracy:
State-of-the-art machine learning architectures
20+ different proprietary machine learning architectures including Higher Order Recurrent Neural Network, as well as CNN, LSTM, Random Forest, and many more for unparalleled accuracy in forecasting future price movements.
Unique voting system
Implementation of Support-Vector Machine (SVM) that turns several neural networks into an ensemble model with the purpose of determining high-probability scenarios.
Rigorous optimization process
Various customized layers for regularization, cross-validation and sensitivity analysis. These are techniques that prevent overfitting (e.g. neuron dropout and pruning), relentlessly tests models subject to training over different time periods, and analyses sensitivity of models to small price variations – all in order to achieve optimized models.