My name is Charles Dana, and I am the CEO of Algorithme.ai : French Applied Math Lab Specialized in Tabular problem classification.
We have developed a python plain text library generator from a .csv : namely "BlackSwan ".
Kaggle | laotse/credit-risk-dataset
We worked on a public dataset of 30 000 people, taking a look a <loan_status>, a binary column that tells if the user developed a credit risk default. By nature the market of classification algorithms has been saturated with Random Forest and Gradient Boosting, which combined with a SHAP-analysis can offer various insights as of the why someone should be at credit risk.
But at Algorithme.ai, we believe in the NP-complete structure of SAT instances as pivot point for classification (and regression) purposes. We generated two python scripts, one with 15 models, one with 90 models, that aim at offering source code for Classification Task.
I've heard that this place is an amazing entry point for feedback, so feel free to deep dive onto our technology available on GitHub AlgorithmeAi.
This innovative approach generates a table of lookalikes with integer ponderation, and creates a confidence% from the ratio #True / #ALL.
This ratio performed well, when presented with Gradient Boosting and Random Forest.
Metric BlackSwan Random Forest Gradient Boosting
ROC AUC 0.901 0.888 0.919
Precision‑Recall AUC 0.821 0.543 0.822
Recall AUC 0.551 0.308 0.632
F1‑score AUC 0.570 0.273 0.656
Youden’s J (max) 0.60 0.56 0.62
Calibration (Brier) 0.15 0.18 0.14
But the core features of this python script lies in the core functions offered by the library.
If you happen to have a tabular problem for classification that would benefit from the BlackSwan insights, feel free to send it to my pro e-mail: charles@algorithme.ai.
Plain text python library to raise the performances of explainable Supervised Learning Classifiers.
Works amazing with LLMs, if you want to give it a go onto an agent, feel free to do so.
Best regards,
Charles
Président, Algorithme SAS (Algorithme.ai)
charles@algorithme.ai
notjustcharles•6mo ago
My name is Charles Dana, and I am the CEO of Algorithme.ai : French Applied Math Lab Specialized in Tabular problem classification.
We have developed a python plain text library generator from a .csv : namely "BlackSwan ".
Kaggle | laotse/credit-risk-dataset
We worked on a public dataset of 30 000 people, taking a look a <loan_status>, a binary column that tells if the user developed a credit risk default. By nature the market of classification algorithms has been saturated with Random Forest and Gradient Boosting, which combined with a SHAP-analysis can offer various insights as of the why someone should be at credit risk.
But at Algorithme.ai, we believe in the NP-complete structure of SAT instances as pivot point for classification (and regression) purposes. We generated two python scripts, one with 15 models, one with 90 models, that aim at offering source code for Classification Task.
I've heard that this place is an amazing entry point for feedback, so feel free to deep dive onto our technology available on GitHub AlgorithmeAi.
This innovative approach generates a table of lookalikes with integer ponderation, and creates a confidence% from the ratio #True / #ALL.
This ratio performed well, when presented with Gradient Boosting and Random Forest.
Metric BlackSwan Random Forest Gradient Boosting
ROC AUC 0.901 0.888 0.919
Precision‑Recall AUC 0.821 0.543 0.822
Recall AUC 0.551 0.308 0.632
F1‑score AUC 0.570 0.273 0.656
Youden’s J (max) 0.60 0.56 0.62
Calibration (Brier) 0.15 0.18 0.14
But the core features of this python script lies in the core functions offered by the library.
If you happen to have a tabular problem for classification that would benefit from the BlackSwan insights, feel free to send it to my pro e-mail: charles@algorithme.ai.
Plain text python library to raise the performances of explainable Supervised Learning Classifiers. Works amazing with LLMs, if you want to give it a go onto an agent, feel free to do so.
Best regards, Charles Président, Algorithme SAS (Algorithme.ai) charles@algorithme.ai