Using AI for item valuation in the insurance industry

One important yet difficult task for insurance companies is to determine what different items are worth. The valuation itself could be a part in deciding how much an insurance should cost, or be used when an item is lost or damaged.

... what is it worth to you?

Want to replicate this use case in your business? Get in touch

02/ The problem

The insurance industry is an industry that is highly dependent on being accurate and concise. For a customer looking to insure a precious item, trust in a factual and correct valuation of it is of utter importance. However, valuations can come with difficulties when the parameters to consider are many, or when information is lacking. By leveraging deep learning in valuation processes, insurance companies can efficiently and instantly analyze images, details provided and combine those with historical data of past valuations.

03/ The opportunity for deep learning

Deep learning can summarize and find patterns in data that are difficult to know or not visible for the naked eye. It also opens up the possibility of combining and drawing conclusions from different types of information, for example combining images and tabular data. Once trained and ready, a deep learning model will be able to pinpoint what it should focus on in incoming images and tabular data in order to provide the right result.

04/ Platform model to use

Depending on what needs to be valued and how you want to do the valuation, we suggest using either tabular data regression, image regression or a combination of the two.

05/ How does the model work?

As with all modern deep learning, it is needed to experiment a little bit to find the optimal solution for the problem in mind. We will focus on using both data sources. To understand and extract information from the images, a convolutional neural network will be used and to find relevant features in the tabular data, a fully-connected neural network will be used. Both these models are available as pre-made modules in the Peltarion platform.

Usually combining different networks for different data types is a challenging task only for advanced users within deep learning. Thankfully with Peltarions intuitive interface, it can be done with no programming experience in a couple of seconds.

06/ Data requirements

Historical data with images and their real valuation is necessary to train the model.

How many items are necessary is hard to know without trying, because it depends on the task and how diverse the data points are. A good starting point is to have a table with at least 100 data points.

07/ Model performance and success

You will see the model performance in the evaluation view on the platform after you have run your first experiment. In these kinds of regression experiments, it is important to look at the average loss as well as the standard deviation. You will learn more about this in the tutorial under useful resources below and if you encounter any obstacles along the way, we at Peltarion are always here to support you.

08/ Where to learn more

If you want to build this yourself with the Peltarion platform. A suggestion would be to follow the tutorial listed below on how to predict real estate prices with tabular data and images. After you have gone through the tutorial, you can use your own dataset in a similar way as described. The tutorial also goes into the details on how to measure model performance and what to do to improve it even further.

If you want to know more about these types of solutions, please see our dedicated pages for image regression and tabular data regression.