Skip to Content

Machine Learning and Tax, Part Deux

Robot learning using basic tools

I was talking to a friend of mine who studies machine learning about our new Taxonics product. During our conversation, he mentioned that my original post on machine learning could stand an update given the Taxonics launch. Instead, I asked him to write something that might help everyone understand machine learning a little better.

Without further adieu, meet Brian Pendleton, a business school classmate of mine who’s currently working on a doctorate in machine learning. I now cede the floor… I mean, post.

Clear as Mud

Thanks for the intro, Ross. Machine Learning sounds scary, which is why I love opportunities to demystify it.

Let me start with some phraseology. You will often hear the terms Artificial Intelligence (AI), Machine Learning (ML) and data science used interchangeably. Data science and AI are separate, but related, fields. ML is a component of AI. Crystal clear, right? No? Let’s break it down.

Data science’s goal is to find knowledge or value in a dataset. Professional sports teams use it to build better teams. Businesses use it to find ways to improve their product offering. Medical researchers use it to track the spread of COVID-19 across the world. The insights offered by using data science allow people to make better decisions than they would without it.

Artificial Intelligence is hard to define with any kind of consensus within the field. At its most basic, it is training an agent of some kind to make similar decisions that a human would given the same information and situation. What’s an agent? It’s a program or a system run by programs. How do we teach one to make decisions for us? Machine learning.

Machine learning is a subfield of AI. It makes up the methods we use to teach the models an agent will use about the world – or at least the part of the world that we want help with. Sometimes you’ll hear a term called Deep Learning. This is simply a specific type of training algorithm within machine learning. Using ML, we can create predictive models from our collected data to help us answer questions before we’ve even asked them.

Let’s Talk About Data.

You’ve probably heard the term big data. However, that name is a little misleading. It’s not the amount of data you have that makes the difference, it’s the relevance of the data. As an example, if you want to know the biking habits of people in the D.C. area, you wouldn’t include data from New York, LA, or Seattle (more data isn’t necessarily better data). There are some problems within the AI space that need a lot of data to increase accuracy, but not every problem needs billions, or even millions of data points.

Finally, you’re probably saying to yourself, “wait that sounds like data science.” And you’re not wrong. Data scientists use machine learning, among other techniques, to do their work.

Why it Matters for Commercial Real Estate

Why do you care about all of this? Other than the fact that machine learning, whether as a part of data science or artificial intelligence, is pervasive throughout society today, it’s also how Cavalry, using Taxonics, is going to help change the industry.

For the how, let’s go back to data. Real estate data is out there, but it’s hard to harness. Here are two examples of how Taxonics (and ML) will make more effective use of the trove of available information.

First, appeals. Taxonics will allow Calvary to identify the best comparable properties to drive tax appeal success (separating the relevant data from irrelevant), getting smarter with each appeal thanks to the machine learning we talked about earlier.

Second, valuations. Thanks to zoning, location, accessibility, etc., two similar properties within the same jurisdiction may have wildly different values. Taxonics’ machine learning capability allows Cavalry to spot valuation trends within – and across – each jurisdiction their clients operate in for more accurate valuations.

While this doesn’t remove people from the equation, it does allow them to operate far more efficiently. In the case of Cavalry and Taxonics, it’s the secret efficiency sauce that allows them to do better, more accurate work while charging less than traditional tax providers.

Back to top