How We Work
The RISE approach is a powerful symbiosis of science, financial and behavioral theory. We leverage AI and machine learning to constantly learn, find, produce, test and refine algorithms and investment strategies, thousands per month.
A machine learning powered approach to automate the process of generating trading ideas allows a far greater scale of ideas to be generated and tested in a short amount of time, a vast departure from the unpredictable time spans and resource-intensive coding of other approaches.
The RISE engine is implemented by an interdisciplinary team of trading professionals, scientists and high-tech experts who constantly research and develop new ways of making RISE’s systems more competitive.
Human / machine hybrid trading strategies
Rise is concerned with sophisticated trading strategies that augment human domain expertise with machine learning techniques. Once these parameters are set, they are then optimized and validated by machine learning techniques which are applied to the human-created vectors in order to create a model, which is back tested using trading algorithms.
The pure AI approach means that the system receives raw data and identifies relations and patterns as well as trading methodologies in order to create profitable strategies. In other words, the AI system identifies both trading methodologies / strategies and generates trading signals. Automatic generation of trading strategies is a much more complex process than automated parameter tuning, but with sufficient resources, Rise’s proprietary processes are more predictable, controllable and efficient.
Predictive and prescriptive models for AI in Trading
Rise is concerned with sophisticated trading strategies that augment human domain expertise with machine learning techniques. Once these parameters are set, they are then optimized and validated by machine learning techniques. Machine learning techniques are applied to the human-created vectors in order to create a model, which is back tested using trading algorithms.
Implementing neural networks for trading
Independently of the choice between predictive and prescriptive models, the structure of the models must be properly chosen in order to achieve the desired general characteristics of the trading strategies. The simplest classes of models that operate on sequences consist of autoregressive models. Within the autoregressive models, the immediate prediction or action—such a buy, sell, or do nothing—is modeled as a function of a fixed number of preceding observations. For example, in the simplest case, the final output for a time series can be generated by looping over all the provided information vectors and using them together with the time series encoding vector as inputs to a neural network.
Validating the Quality of Trading Systems
Validating the Quality of Trading Systems Trading systems must be cross-validated and checked for overfitting. If not done correctly, this is the spot where the most dangerous errors can be made. The core aspect of overfitting is when the model starts to «memorize» the data, instead of finding patterns. The solution is to bind the optimizable parameters of the strategy and thus reduce degrees of freedom.
Trading strategy development
All Rise-generated trading strategies have a very low correlation with each other. Most of the new models start with a simple idea. The next step is to verify that a trading idea has potential and research it further. The process can take quite some time as each trading idea differs in terms of where to look for deviations and anomalies in price or behavioural patterns. Rise’s AI and machine learning systems speed up the process to help validate or dismiss the initial idea.
Performance and risk data for each of Rise’s strategies is the feed for our portfolio optimization. Additionally, all trading systems are developed to produce returns that do not correlate with one another. The ideal portfolio is selected depending on maximum Sharpe ratio or minimum volatility along the «efficient frontier» as shown in the image below.
The importance of risk management
In Rise’s view, effective risk management begins with the design of the trading strategy itself. The model has, therefore, a substantial predicting power. Rise’s more traditional methods for risk management are based on quantitative analysis of market data and behavioural finance. Another key aspect of Rise’s approach to risk management is transparency toward clients. augment human domain expertise with machine learning techniques. Once these parameters are set, they are then optimized and validated by machine learning techniques. Machine learning techniques are applied to the human-created vectors in order to create a model, which is back tested using trading algorithms.
Want to know more? Sign-up for our Newsletter.
Sign up for our Newsletter to get news & updates about our products, strategies and releases.