Causal AI understands, explains and predicts financial markets, offering the next generation of portfolio optimization and risk management capability, enabling asset managers to unlock latent market value.

Modern challenges in capital markets are well known
Investors demand lower fees and greater explainability from discretionary managers, while systematic funds must constantly extract new signals from an ocean of marginally useful data. Asset managers can turn to Machine Learning to stay on top of the problem, but current AI techniques are often inadequate:

• Basic AutoML techniques are widely used, and their capacity to produce alpha has diminished
• Non-causal ML techniques based on pairwise relationships will necessarily overfit to the past, leading to high risk and low understanding around future regime changes and unexpected events. 
• They fail to translate predictions and models into actionable trading decisions.

Solutions

• Trading Signals
  causalNet is a state-of-the-art modelling system that extracts the causal drivers from large datasets and marries them with human domain knowledge in one single streamlined process. The resulting constrained neural network outperforms non-causal AutoML models in out-of-sample testing on a huge range of forecasting tasks, ready to feed into any trading strategy.
• Factor Investing
  Causal Discovery uncovers the driving forces behind cross-sectional returns, allowing for the robust construction of latent variables for factor analysis as well as the discovery of new factors beyond those found by standard correlation or PCA techniques.

  Using causalNet, we can construct factor portfolios that systematically control for unintended exposures — removing cross-correlation effects between factors and dynamically constructing factors based on data, rather than imposing arbitrary definitions.

• Risk Management
  Our Causal Discovery methods can identify the true causal drivers of exposure for a portfolio, rather than focusing on spurious correlations. Using causalNet, we can generate realistic scenarios that capture fundamental regime shifts during periods of market stress while also remaining intuitive and explicable.  
• Evaluating Datasets
  In today’s market, having the right data is a key to success. Evaluating the quality of data is nonetheless a challenging task. With causaLens this task is made easier with our proprietary dataset evaluator: benchmark the candidate data against public datasets and quantify the impact of the data on your KPIs, while avoiding the spurious correlations in all large datasets.
• Portfolio Optimization
  Portfolio construction methods based on correlation matrices impose penalties on spurious factors, limiting performance. causaLens has pioneered Causal Hierarchical Risk Parity, a portfolio optimization model based on robust causal clustering techniques that produces lower volatility and higher risk-adjusted returns than the traditional techniques.
• Signal Generation
  Poor input signals will lead to poor results in any forecasting task. The causaLens platform feature engineering pipeline generates a fully customizable set of features while automatically controlling for redundancy and statistical significance, greatly enhancing the search space while avoiding the problems of exponential runaway.