In a previous post, I promised to elaborate on why a “safety mechanism” is essential when Machine Learning (e.g. Deep Learning) is used to automate decision making.

In a classic computer software situation, the computer executes code written by a programmer in a chosen programming language. As such, the software behaves in a very predictable manner. However, when Deep Learning is used, the decision is not made by an explicit programmer’s code. Instead, it is made by a pre-trained model.

One might ask, why not “debug” the model when things go wrong? It’s a computer code at the end of the day!

Unfortunately, while Deep Learning might appear simple on paper (see image on the left), this is just a simplified abstraction. In reality, it’s totally different (see image on the right). It should be clear by now why AI is black box and no one can really tell where the error may be.

For this reason, it makes sense to test the output against well-know “good” limits and take action when the prediction is off range. For example, override the prediction or notify the user.

Can it Be Done?

Spoiler alert: yes it can.

A couple of years ago, I started digging around about using AI for valuing residential properties. Back then, everyone said it was too complicated for “algorithms” to handle, and AVMs only served to lure customers. I challenged that and built AccuVal, which has proven that it’s possible to get 90% of the valuation done under 30 seconds with unprecedented accuracy.

The same scenario has repeated itself for valuing commercial properties. The graph in this post illustrates the result of a very early experimental AI model that was able to value 10,000 commercial properties (that it didn’t see during training) with an average accuracy ranging between 70% and 85% (for challenging areas and stable areas accordingly).

I will be posting more as the development progresses so make sure to watch this space if this interests you 😉

The common practice in valuing residential properties is to find the prices of similar properties sold recently. Too recent could mean there may be none (or insufficient), while going back too far isn’t particularly a good idea.

So how did I address these challenges in AccuVal?

By applying the appropriate inflation to historical transactions, we are effectively fast-forwarding them to the current time. For example, if a property was sold in 2017 for £250K, all things being equal, it would be as if it was sold for about 269K in 2021.

What’s left is to predict the short-term changes (30 days) to property prices and apply that to the property being valued today, so by the time the property is listed and viewed, it’s valuation would be as current as possible.

AccuVal is a free valuation web app that uses pure Machine Learning to value residential properties in England and Wales, including properties that don’t yet exist or recently changed.

Residential property valuation relies heavily on properties sold nearby. In other words, relying on the past. However, property prices change all the time. Therefore, valuations based on historical records need to be adjusted (up or down) according to the current, or even better, short-term market trends. This involves predicting the change before it has happened!

Using various datasets, I trained a simple ML model to predict price change and plotted the results on a map.

According to the prediction, the change ranges between -10% and +11% across the UK.

Once the ML model is verified, I will incorporate it into AccuVal. AccuVal is an AI based automated valuation tool that’s among the best in the UK, and it’s completely free to use.

Imagine your dream detached house right at Trafalgar Square. How much would it worth?

Unsurprisingly, it turned out that no detached houses were sold there. Therefore, there is nothing to compare with.

Nevertheless, I assumed an “imaginary” house that’s 10 rooms and 4000sqrtFT and used AccuVal to get a valuation, and bam! I got £5,015,000!

I have no idea if that was a reasonable valuation, but it’s nice to have a tool that can be used even in these situations.