What Every Real Estate Professional Needs to Know About Ensemble Modeling
From The Buxton Co
By Adrian Harvey
Staying on top of the latest buzzwords and trends is a challenge in any profession, and the real estate industry is no exception.
Today’s real estate professionals researching site selection consultants and tools may encounter the term “ensemble modeling,” which may be unfamiliar. In this blog, we’ll explore what ensemble modeling is, how it relates to more familiar concepts like regression modeling, and the scenarios where ensemble modeling is most applicable.
What is ensemble modeling?
In ensemble modeling, multiple models are employed to answer a specific question. In a real estate scenario, the question could be something such as, “what’s the expected performance of this site?” Each model takes an independent approach to answer the question and may use different modeling types (linear regression, decision tree, etc.), different model classifications (breakouts), different variables, etc. The results are determined based on either a single best answer or a combination of answers from the multiple models.
How is ensemble modeling different from regression modeling?
Ensemble modeling is not a “modeling” approach in the same sense as regression or logical regression modeling. Instead, it is simply the practice of using multiple models or modeling techniques.
Ensemble modeling isn’t a substitute for regression modeling because it may use regression modeling to arrive at an answer. The difference is simply the number of models involved in trying to answer the question.
Is ensemble modeling a new approach?
Ensemble modeling has been used in real estate for many years, although the term itself has only recently become familiar. As processing capabilities have increased and data has become more accessible, ensemble modeling has become more common.
When should ensemble modeling be used?
Ensemble modeling is commonly used in scenarios where a single modeling technique may be inadequate to account for the wide range of factors that influence an outcome (e.g., predicting climate changes over time). At Buxton, we use ensemble modeling when we’re analyzing complex real estate scenarios, such as those involving differences in population density, regions and operations or factors with non-linear relationships.
While ensemble modeling provides the benefit of multiple perspectives, there are also drawbacks. Having multiple perspectives may affect the interpretability of the analysis since the results may disagree – making the output confusing to interpret and act on. What usually ends up happening is that the output becomes “black box” to make it more useful. Instead of seeing the full range of model responses, you simply see the final “vote.”
Is ensemble modeling the right approach for my business?
Ensemble modeling comes in many forms depending on the scenario. A good analytics partner will seek to understand your situation and goals before applying the modeling approach that makes the most sense, rather than always using a specific modeling technique.
Buxton is the leading customer analytics firm that helps organizations identify who their customers are, where those customers are located, and the value those customers have to the organization.
2651 South Polaris Drive
Fort Worth, TX 76137
If you have an opinion on the retailing or retail real estate industries, take this opportunity to share your thoughts. Articles should run between 400 and 800 words. Topics can, be general in nature, consumer observation or specific to retail concepts or practices.
Articles will be posted for at least one week and will then be placed in the Editorial Archives. All articles submitted will be read and considered but we cannot guarantee publication. Each published article will carry the submitters byline (if desired) and is a free service to our community.
Article ideas and suggestions are also always welcomed. Contact PVS@PlainVanillaShell.com