Benchmark: xtractis vs. RANDOM FORESTS, boosted trees & NEURAL NETWORKS

The objective of this study is to instantly detect, without the aid of a camera, radar, or lidar, if a vehicle is in an emergency situation in order to automatically activate the emergency braking and trigger the ABS.

It illustrates the ability of the Trustworthy AI xtractis to automatically induce knowledge in the form of predictive and intelligible mathematical relationships to model the complex decision of emergency braking from the automatic analysis of the driving situation. In terms of road safety, the slightest mistake can prove fatal, so it is necessary to approach the perfect model.

In the end, xtractis generates a classification model composed of 25 unchained decisional rules, using 12 predictors among the 17 potentials, and predicts the state of the emergency for the infinity of the points of the decision space with proven reliability.


– Modeling type and reference data

– Automatic induction xtractis process

– Top-IVE: Best predictive and intelligible model

– Top-IVE performance

– Refusal of decision

– Intelligibility of the model and Explainability of the decision

– Example: Prediction for the driving situation BRUNO_FREIN103.TXT_105.97_659660

– Resources of the xtractis process (Induction + Validation + Deduction)

– Benchmark xtractis versus Random forest, Boosted trees & Neural networks

– “Classic” Fuzzy AI versus Augmented Fuzzy AI xtractis

– Conclusion and Benefits of the Trustworthy AI xtractis

Results by xtractis® GENERATE 11.3.40047 (November 2021)
First use-case version: February 2022
Current Version: February 2022 (v1.0)