Damage Prediction

Pattern detection on images

Humans don't face issues to read text on images. The image consists of thousands of pixels in a different colors. The human eye can detect the pattern of the text in it.

It's very difficult to develop a classical computer program with the same capabilities. In this example one could delete all blue and white pixels and try to compare the remaining characters with a font. This program wouldn't work for blue text. With the help of Deep Learning this challenge can be solved.

Pattern detection in data

In other data like assurance data it's very difficult for the human eye to detect pattern.

Age Sex HP Color Damage
45 M 60 white 310
20 F 80 blue 0
33 M 70 black 500
34 F 250 black 2000
... ... ... ... ...

In this data could be hidden that there is a damage peek for 50-60-agers driving high-powered cars. This can't be detected with the human eye in many records having many attributes.

A Deep Learning software could solve this challenge.

How can the damage costs be forecasted?

You can develop a neural network that takes all attributes as inouts an the damage cost as output. If you train the network with assurance contract data and the damage costs of past you could forecast the damage costs already before signing the contract.