Authored by: Matthew Erasmus – Director of Projects and Engineering at FALX
“AI is not just a trend in real estate; it’s a fundamental shift that will redefine the industry, unlocking new opportunities and efficiencies.” – Ryan Williams, CEO of Cadre
The COVID-19 pandemic in 2020 catalysed the need for innovative virtual systems in the economy, as businesses were compelled to adapt to a new era of social distancing and remote work. This period sparked a wave of creative thinking, leading to unconventional ideas that have reshaped traditional practices across numerous industries. The application of AI, particularly machine learning (ML), has presented remarkable opportunities for enhancing various aspects of property valuation and municipal taxation. The lingering question is whether the industry will embrace this new technology as the standard for the future.
Machine Learning Unveiled
Defined as a branch of AI that empowers computers to learn from data and improve their performance over time without the use of explicitly defined numerical functions — colloquially known as the mightiest of all universal function approximators. ML struts its feathers when presented with complex, nonlinear problems that consist of large, multi-dimensional datasets. ML has the remarkable ability to incorporate nuance and pick up complex patterns in data, often surpassing the abilities of experienced humans.
Machine Learning in Mass Appraisal
The vast amounts of data and detailed property attributes encountered in the mass appraisal process make it an ideal candidate for machine learning techniques. Among these, gradient boosted machines have emerged as the worldwide favourite architecture. According to the IPTI, professional valuers have observed sustained or improved accuracies compared to traditional methods, with efficiency improvements of up to 85% in terms of man-hours for the valuation process. Moreover, better customer satisfaction is evidenced by reduced numbers of appeals and inquiries. The success of these systems is reliant on good quality data but their implementation has provided early adopters with a strategic advantage, allowing their professional resources to focus on vital market research and stakeholder engagement.
Digitising the Past with Optical Character Recognition (OCR)
A wealth of information resides in paper-based archives, yet these files remain largely inaccessible beyond hours of manual ingestion. Enter OCR, a method for autonomously digitizing scanned documents through text recognition. These techniques are capable of enriching and completing databases that previously did not exist in the digital sphere.
Convolutional Neural Networks outling Property Detail
Image recognition – a brilliant example of a classification problem. From self-driving cars to Facebook tagging; convolutional neural networks have proven their abilities to perform these tasks seamlessly. CNNs provide considerable time-saving opportunities when it comes to the extraction and digitization of land use from aerial photography and property volume, as well as finer details from LiDAR imaging. LiDAR is a point cloud system that constructs a 3D model using laser technology. This information, when parsed to a CNN, provides the opportunity to accurately identify external property features – previously done by manually scanning every property.
As we journey towards a future shaped by technological innovation, it becomes evident that these advancements harbour immense potential to revolutionise the realms of property governance and taxation. With precise and efficient valuation methods coupled with enriched data sources that evolve alongside our dynamic landscapes. Picture a world where our datasets mirror reality, seamlessly reflecting the latest sales or newest renovations. This is the future.