| CPC G06F 16/29 (2019.01) [G06F 16/24573 (2019.01); G06F 16/2465 (2019.01); G06F 16/248 (2019.01); G06F 16/288 (2019.01); G06N 20/00 (2019.01); G06Q 50/16 (2013.01)] | 20 Claims |
|
1. A method comprising:
by processing circuitry, for each respective real estate property of a plurality of real estate properties,
a) identifying at least one geospatial feature of a plurality of geospatial features located within a predetermined distance of a location of the respective real estate property,
b) calculating, for each respective geospatial feature of the at least one geospatial feature, a distance between the location of the respective real estate property and the respective geospatial feature, and
c) storing, to a non-transitory storage region, a feature vector for the respective real estate property, the feature vector comprising
property features comprising at least a subset of a plurality of property attributes, and
the at least one geospatial feature including, for each respective geospatial feature of the at least one geospatial feature, the distance,
wherein the storing results in a plurality of feature vectors being stored to the non-transitory storage region;
by the processing circuitry, for each respective subset of a plurality of subsets of the plurality of real estate properties, evaluating, using one or more machine learning algorithms, a respective portion of the plurality of feature vectors corresponding to the subset of real estate properties to identify a respective set of impactful features identified as having a significant impact on property value for the respective subset of real estate properties, wherein
the respective subset of real estate properties share at least one of a property type, a target geographic region, a target demographic group, or a same geospatial feature of the plurality of geospatial features;
training, by the processing circuitry, a plurality of machine learning data models, wherein
each trained data model of the plurality of machine learning data models is configured to, using a respective portion of the plurality of geospatial features and a respective portion of the plurality of property attributes corresponding to a given set of impactful features, apply machine learning to the feature vectors of at least a portion of the plurality of real estate properties to identify one or more of a) a set of similar real estate properties determined to be similar based at least in part on the given set of impactful features or b) a property valuation for each real estate property of one or more real estate properties determined to be similar based at least in part on the given set of impactful features; and
by the processing circuitry, responsive a request submitted by a remote computing device,
i) identifying a set of feature data for the request, the set of feature data comprising
one or more geospatial features, and
at least one property attribute of the plurality of property attributes,
ii) selecting a trained data model of the plurality of machine learning data models,
iii) obtaining, responsive to the trained data model applying the set of feature data, result information, and
iv) providing, to the remote computing device, the result information.
|