Valeri identifies valuable market patterns and property insights through the application of machine learning and Artificial Intelligence. Valeri is constantly analysing city-specific data sets, including sales data and unique proprietary layers, to learn patterns and predict land and property values with industry leading accuracy
What sets Valeri apart is its ability to enhance the analytical process by converting difficult to value qualitative factors such as planning, property and market attributes into an equivalent dollar value.
This is how we do it:
No two locations will have the exact same property insights and characteristics. Each city, suburb and even street has its own unique tastes, trends and market characteristics, which is why Valeri’s systems are rigorously trained with curated data learning sets, specific to the city, suburb and street level.
PointData has developed systems to carefully cleanse proprietary data sets for false outliers, mis-entered values and poor or missing features to ensure our machine learning algorithms are trained on good quality data.
PointData has developed a unique process to dynamically grow the area of influence surrounding each property, irrespective of arbitrary boundary and suburb constraints. Valeri uses up to 2000 comparable sales to achieve a statistically significant sample and only like for like properties and sales.
Valeri’s growth indices are calculated separately for property and land, in both space (spatially) and time (temporally), at a detailed neighbourhood level.
PointData’s AVMs are powered by Valeri’s Artificial Intelligence systems, which are consistently processing proprietary data learning sets under close supervision of our data scientists. Valeri uses machine learning to relate like for like properties and location features in multi-dimensional space.
PointData has developed a system that mimics best practice out of sample testing, usually applied by banks to test the validity and accuracy of an AVM. PointData’s Forecast Standard Deviation (FSD) is therefore far more representative of the real error and price range compared to traditional in-sample methods.