The Neuravest process converts raw, unstructured data into meaningful data points that can be analyzed through machine learning and deployed into differentiated investment strategies. By applying our data science expertise, we can extract specific features from a variety of data sets and normalize them so that they co-exist in a multi-factor investment model. Once we’ve completed the feature engineering, we can automate our machine learning process and begin portfolio model building.
By using supervised machine learning, Neuravest goes back in time to find specific historical outcomes to optimize for our investment strategies. We can then extract which data sets and other quantitative factors contributed most directly to deliver these outcomes. Our platform analyzes millions of combinations of data sets and factors to determine how they work together to get a sustainable (and repeatable) outcome. From here, we generate and deploy multi-factor portfolio models.
Before deploying any client investment strategy, Neuravest uses an objective backtesting process to validate impact and quantify potential added value.
This proven methodology includes:
- Empirical simulation of the signal-to-trade efficacy by rolling back time and trading based on point-in-time data.
- Out-of-sample trading simulation that is free of any selection-survivor or look-ahead bias.
- Integration of transaction costs and market impact analysis.
- Complete performance attribution analysis including the transaction audit trail.
Live Broker Integration
The Neuravest platform seamlessly integrates with a client’s destination account via Financial Information Exchange (FIX) and our application programming interface (API). We replicate our models through synchronizing the allocations between our platform and the destination portfolio. This automatically generates trades and determines when to buy, sell, and rebalance. Our platform is broker-neutral, enabling our clients to execute orders using their own trading tools and strategies.