Ml for matching
We already did some experiment to with different ML-Model to improve our matching proposals. Out of the different algorithm until now, random forest seems to be the best fit to automatically find correct candidates. The main problem are the very few inputs.
The goal of the ML-Support for matchings is to do less work. We can achieve this goal by reducing the amount of bad matching proposals. The model seems not very bad, but its not good enough to do automatching. To get to this point we need more data for each person.
-
Bring the current experimental ML into production -
Add a alpha preview feature to test the ML with real data