The solution

Stand Up Meeting

Deepaiy predicts the number of days a credit is exposed.

The impact of accurate predictions on human decisions has a great extent: for example lower cost of funding, lower NPE rates, improved pricing of risk, earlier warning signals, accelerated collection.

Building
Blocks

01

Data Collection

Deepaiy AI/ML proprietary models process the debtor cash-flow enriched with small context information and public data. The data mart is segregated and fully owned by the client.

03

Advanced Analytics

In each run several AI/ML algorithms compete to perform the most accurate predictions. The top performer is chosen based on six back testing metrics. 

02

Data Pooling

The client joins the Deepaiy Data Pool (DDP) to overcome the scarcity of data. The data are collected anonymously and the entity is masked to safeguard competition.

04

Business Intelligence

Deepaiy predictions can be embedded in client's systems or published on the web Business Intelligence interface, personalised with client's metrics.

Why is Deepaiy Data Pool (DDP) different?

1. Small Data Issue

  • The clients with small data join the Deepaiy Data Pool (DDP)

  • The data are submitted at entity or business line level

  • Before sending the data, the debtors are masked and the entity is anonymized 

2. How DDP works

  • A and B are both creditor of D and join DDP

  • D doesn't fulfill the obligation with A

  • In a standard data pool, B has an advantage thanks to the information submitted by A

  • In DDP D is masked and the creditor A is anonymized, so B doesn't have a direct advantage

3. Central Agent

  • AI/ML proprietary models are trained on the central DDP

  • The trained models are called Central Agents

  • The client data is submitted to the Central Agents to perform more accurate predictions