AI & Real Estate: Part 1
More Data, More Intelligence
While the term Artificial Intelligence has existed for decades, the concept has taken a completely new direction in only the last few years. The latest wave of PropTech disruption started almost five years ago with the arrival of machine learning and the early stages of predictive analytics.
As companies race to harness these technologies to remain current and tap into statistical analysis like never before, new algorithmic advances are positioned to replace the siloed datasets that have accumulated over the past 20 years on government websites or large real estate data platforms.
While the AI advances in this sector are surely changing the current PropTech landscape substantially, this revolution is proving to be a quiet one that will not be represented in products and volume of data, but it will disrupt and renew the intelligence and thought behind existing products and services.
Real estate has historically lagged behind other industries in terms of technological advancements and new concepts, and while this has proven to be true with AI as well, it is finally coming around thanks to advancements and avenues made possible with machine learning and neural networks.
What is Machine Learning?
To understand the capabilities of predictive analytics in the commercial real estate market, it is important to grasp some sense of the kind of technology at stake. Tools used are primarily based on tried and true statistical methods that are applied in more sophisticated systems. The widespread usage and increased educational material around machine learning, deep learning, or neural networks allows the new generation of PropTech startups to be able to make statistical models at a higher level.
Machine learning is the development of programs that can change when exposed to new data. The process of machine learning is similar to that of data mining, as both systems search and scrape through data to look for patterns. Where they differ is that instead of extracting data for human comprehension, machine learning uses these patterns in data to adjust program actions automatically.
This means that multiple regression models used to forecast something like rental prices can be constantly re-estimated in real time as new data enters the system.
But…how do you teach a machine?
All of the aforementioned advanced statistical methods follow very similar learning processes. First, a “training phase” takes place where the machine is fed datasets and, effectively, “learns”. In other words, it absorbs the datasets complexities and tries to weigh each factor (i.e. house characteristics) on the output value (i.e. housing prices).
Following this is a “testing phase”, in which the machine previously trained is tested against a set of data where the output is known. This allows us to observe and track how accurately the algorithm predicts the output value. This is the “prediction phase” where we use the algorithm to guess the output value of a dataset with an unknown output value.
How is it used in Real Estate?
Thus far, leading companies in PropTech have invested a great deal of effort, time and money into stabilizing simple regression models to complete short term analysis. The predictive precision that is possible by machine learning is pushing the boundaries of forecasting entirely. From just a few months to a 10 year span, the ability to predict the future of deals, markets, and pricing is drastically changing the perspectives of investors.
Prediction is taking decision making to a whole new stage, and in real estate markets, this is changing the way people buy and sell properties as we speak. By analyzing the statistical meaning of the data at hand, the real estate industry is now capable to see the probability of a deal’s success.
With predictive capabilities opening new horizons to urban development and real estate, data integrity is the nucleus of keeping it all together. By leaning cleaner, more accurate datasets and constantly training and refining algorithms, certain platforms will be able to predict prices with greater accuracy and detail over time.
Current Limitations
Although forecasting should soon be game changing for the real estate industry, most companies in this field are still holding on to reputations based on more antiquated models of the industry. The traditional approach of risk analysis (spending almost unlimited resources on feasibility studies) is still recognized as standard. Most investors and real estate players alike trust the old system and old processes, and react negatively to tech-evangelism.
Maybe the most common issue in predictive analytics is the integrity of the datasets themselves. If large databases have been aggregated over time, it is often hard to judge the quality of the data.
Also, users are sometimes invited by certain platforms to claim their own property and enter data themselves — data that is then factored into these important predictions. This can leave the data outdated, skewed, or just plain ol’ incorrect, and results in inaccurate forecasting and predictive models.
Next Time…
In Part 2 of the AI & Real Estate series, we will take a look at companies that are currently using AI in PropTech with contemporary statistical analysis. Stay tuned by following us on Medium here, or Twitter here.
