|
 
|
[Sap16] Predictive versus Generative Modelling: a Challenge for (Social) Sciences
Conférences invitées : DSSR, February 2016, pp.xx, Naples, Italie,Mots clés: black-box models, data science, inference, machine learning
Résumé:
Classical inference corresponds to the role of statistics as an auxiliary of sciences. However in most sciences, a good model should also provides accurate predictions, which becomes the sole criterium in decision sciences like pattern recognition, customer behaviour, etc. The most efficient predictive models are rather black-box algorithms like random forests or deep learning.
The use of black-box models fitted for massive data is probably the main challenge for social sciences due to their lack of interpretability. Getting better predictions, thanks to a better understanding of the real world, needs to combine statistics and machine learning with causal inference.
Commentaires:
Data Science & Social Research Conference University of Napoli Federico II , February 17-19, 2016
BibTeX
| @inproceedings { | |||
| Sap16, | |||
| title | = | "{Predictive versus Generative Modelling: a Challenge for (Social) Sciences}", | |
| author | = | " G. Saporta ", | |
| booktitle | = | "{DSSR}", | |
| year | = | 2016, | |
| month | = | "February", | |
| pages | = | "xx", | |
| address | = | "Naples, Italie", | |
| note | = | "{Data Science & Social Research Conference University of Napoli Federico II , February 17-19, 2016}", | |
| } | |||
