Symbolic
Regression

Why Symbolic Regression?

If we take a system of which we have X input values ​​and Y output values, a typical AI problem will be to predict future Y’s. In complex real situations where Deep Learning solutions are used, the neural network will give a prediction of Y with a good level of accuracy but it will not be possible to explain how this prediction was achieved.

One of the most frequent criticisms leveled at AI and its applications is that, however effective, it remains a Black Box.

Symbolic Regression (RS) can be seen as an unboxing operation: it is a technique that not only allows us to predict future Y’s, but to mathematically derive the process that binds the X’s to the Y’s and therefore why a certain output is obtained. given some input.

For this reason, a good RS returns results that are more interpretable, more generallizable and robust than the classic techniques of AI and Machine Learning.

What is the Symbolic Regression?

Given a set of inputs X and one of output Y, the Symbolic Regression is a type of Regression that identifies the function f (X) = Y, that is the mathematical expression that best fits the dataset both from the point of view of accuracy than the simplicity of the formula.

Unlike the classical techniques of Regression, this approach by its nature does not require radicated assumptions about the type of model and is therefore very useful especially in the exploratory phases or in those situations in which there is no previous knowledge about the characteristics of the phenomenon under study.

Examples of
Application sectors

Fraud detection

With RS it is not only possible to identify a fraudulent action (anomaly) but to derive the most relevant factors to discover a fraudulent action.

Churn Prediction

With RS it is not only possible to predict customer abandon but to formulate and explain it mathematically, thus possibly also providing tools to manage those aspects that most lead a customer to abandon.

Predictive Maintenance

With RS it is not only possible to predict the anomaly in advance (breakdown / malfunction of the machinery) but to identify the factors that led to this breakdown and provide tools to apply targeted and corrective interventions.

Request an online demo

One of our consultants will assist you with the explanation.

By filling out the form above and sending a request for information, I declare that I have read the privacy policy of ATG Artificial Intelligence SRL and accept the terms.