As the adoption of new technology in investment management continues to grow, AI based insights have proven invaluable for investment managers seeking to add additional alpha to their portfolios. However, one deterrent in fully embracing AI is the need to explain AI-based decisions to stakeholders. Fund managers cannot use black box algorithms that don’t give insight into what the machine is thinking.
Transparent, trusted, and easily explainable AI models are a key to success when adopting AI in your business, leading all the way up to a successful AI transformation of your organization and its decision making processes.
Explainability provides visibility and transparency into the machine’s decision-making process, encompassing the entire process from calculations to decision-making. At Boosted.ai, we are driven by the success that our customers achieve in their investment strategies and decisions, aided by Explainable AI. We began our explainability efforts by enabling the Feature Word Cloud – it shows the exact features that went in your model and contributed the most in explaining the decisions that the model made.
Furthering our efforts to provide explainable and transparent artificial intelligence for investment managers, we have just added a more sophisticated explainability module in our AI platform Boosted Insights, called Feature Importance.
The above screenshot from Boosted Insights showcases our Feature Importance module. It shows the feature, its impact on the model’s output (either leaning towards being a stronger buy or a stronger sell) and its overall importance compared to the other features in that model.
Feature Importance generates a series of graphs that highlights how individual features tend to impact the machine’s decisions, explorable over the entire model timeline. To add to this, we have also added the ability to analyze two-feature interactions to uncover more insights, as shown in the image below.
The picture above shows how Analyst Expectations relate to the 9 Month Price Momentum of stock in this model. If Analyst Expectations are really strong and Price Momentum is negative, our machine learning views that as a buy signal. On the other hand, if Analyst Expectations are weak and Price Momentum is positive, then it is a sell signal. This feature will allow users a deeper understanding of what the machine is doing and interesting relationships that it has found across the input space.
To learn more about our Feature Importance in Boosted Insights, please read more here.
To book a personalized demo to see how you can use artificial intelligence to boost your investment strategies, please visit here.