Predictive models development
S2E is developing, with the help of Its partners and Univerity institutes some predictive algorithms based on Learning machine and neural networks.
S2E applies those models to some use cases referring to the banking world and to consumer credit, as well as to other sectors connected to Farma world.
In particular, for a bank customer, S2E has applied some algorithms based on Learning machines and neural networks in order to identify the account holders who will ask for personal credits and the potential ammount requested.
Artificial intelligence applied in Finance area
20.000 cases of funding
70% for training
30% for test
Implemented neural network
Principal advantages of neural network in the case under consideration
- On time prediction (instead of a class as in the classical statistics methods)
- Extreme accuracy
Mistake on the AMMOUNT (differences between real and predicted):
with credit ranging between 1.500,00€ and 75.000,00€ in real data
Mistake on the DURATION (difference between real and predicted):
⇨ 60 days
with duration ranging between 30 and 3.600 days in real data