Systems and methods for generating feature sets for use in entity evaluation

1 . A system for predicting multi-family property rental income strength across geographic areas, the system comprising:

at least one non-transitory computer readable medium configured to store

a plurality of machine learning models, and

a plurality of location feature data sets; and

one or more processors configured to

consolidate the plurality of location feature data sets to the at least one non-transitory computer readable medium, the plurality of location feature data sets comprising, for each respective geographic region of a plurality of geographic regions, a respective location feature data set relevant to the respective geographic region, the respective location feature data set comprising, for each location feature of a plurality of location features, a respective one or more location feature data values, wherein

the plurality of geographic regions is defined at multiple levels of granularity, and

location feature data values corresponding to different location features of the plurality of location features are available at different levels of granularity of the multiple levels of granularity,

transform a plurality of feature data values of a set of location features of the plurality of location features from an original granularity level of the multiple levels of granularity as consolidated in the plurality of location feature data sets to another granularity level of the multiple levels of granularity,

store the transformed plurality of feature data values to the at least one non-transitory computer readable medium as part of the plurality of location feature data sets, wherein

the transforming populates the plurality of location feature data sets with feature data values for each of the set of location features at each granularity level of the multiple levels of granularity,

identify a plurality of output variables comprising, for each respective geographic region of at least a portion of the plurality of geographic regions, a respective output variable representing rental income per unit time in the respective geographic region,

using the plurality of location feature data sets and the plurality of output variables, train, for each respective geographic area of a plurality of geographic areas, at least one respective machine learning model of the plurality of machine learning models to predict i) an average rental income per unit value for multifamily buildings within the respective geographic area, and ii) a set of feature-level scores indicating a contribution strength of each location feature of at least a portion of the plurality of location features to the average rental income per unit value, wherein

the plurality of machine learning models comprise at least one machine learning model for each granularity level of the multiple levels of granularity,

determine, in real time responsive to receiving input via a user interface screen presented to a user of a remote computing device, a location score calculation data set representing at least one geographic region of the plurality of geographic regions and comprising, for each respective geographic region of the at least one geographic region, the corresponding location feature data set of the plurality of location feature data sets to the respective geographic region,

apply the location score calculation data set to one or more machine learning models of the plurality of machine learning models to obtain

a respective average rental income per unit value for multifamily buildings within each geographic region of the at least one geographic region, and

a respective set of feature-level scores within each geographic location region of the at least one geographic region, wherein

the one or more machine learning models are relevant to the at least one geographic region, and

in real-time responsive to receiving the input via the user interface screen, return, for presentation at the user interface screen of the remote computing device,

at least one rental income prediction based on the respective average rental income per unit value for multifamily buildings within each geographic region of the at least one geographic region, and

feature score information corresponding to at least a portion of features of the respective set of feature-level scores corresponding to one or more geographic regions of the at least one geographic region.

2 . The system of claim 1 , wherein each geographic region of the at least one geographic region is defined by a census tract.

3 . The system of claim 1 , wherein the average rental income per unit value is calculated as a rental income strength on a log scale.

4 . The system of claim 3 , wherein the one or more processors are further configured to transform the rental income strength to a linear scale as a normalized location quality score.

5 . The system of claim 1 , wherein transforming the plurality of feature data values comprises, for each respective location feature of a subset of the plurality of location features, transforming the respective one or more location feature data values of each respective location feature data set relevant to at least two overlapping geographic regions of the plurality of geographic regions overlapping with a geographic area to produce a corresponding one or more location feature data values applicable to the geographic area, wherein the geographic area is at a different level of granularity of the multiple levels of granularity than the at least two overlapping geographic regions.

6 . The system of claim 5 , wherein transforming the respective one or more location feature data values comprises applying a residential-area ratio as a weighting factor to calculate a weighted average from the respective one or more location feature data values of the at least two overlapping geographic regions.

7 . The system of claim 1 , wherein returning the feature score information comprises returning, for each respective feature of the portion of features, a respective current metric for the respective feature at each geographic region of the at least one geographic region, and a respective score for the respective feature at each geographic region of the at least one geographic region.

8 . The system of claim 1 , wherein:

the at least one geographic region comprises a set of geographic areas at a smaller level of the multiple levels of granularity, wherein the set of geographic areas comprises all geographic areas at the smaller level located within a selected geographic region defined at a larger level of granularity of the multiple levels of granularity; and

the at least one rental income prediction comprises a percentile score representing a relative ranking of rental income score of a subject geographic area of the set of geographic areas compared to all other geographic areas of the set of geographic areas.

9 . The system of claim 8 , wherein:

the set of geographic areas are census tracts; and

the selected geographic region is a metropolitan statistical area (MSA).

10 . The system of claim 1 , wherein the one or more processors are further configured to:

identify a plurality of missing feature data values of one or more location features of the plurality of location features corresponding to a first geographic area at a first granularity level of the multiple levels of granularity; and

fill in each missing feature data value of the plurality of missing feature data values.

11 . The system of claim 10 , wherein filling in each missing feature data value comprises, for at least a portion of the plurality of missing feature data values:

for each respective location feature of the one or more location features and for each respective feature data value of the portion of the plurality of missing feature data values corresponding to the respective location feature, impute an imputed feature value using one or more feature data values of the respective location feature corresponding to a second geographic area at the first granularity level, wherein the second geographic area is physically adjacent to the first geographic area.

12 . The system of claim 1 , wherein:

the one or more processors are further configured to store, to a non-transitory computer-readable data store, the respective average rental income per unit value for multifamily buildings within each geographic region of the at least one geographic region; and

the at least one rental income prediction comprises a trended location score generated using past average rental income per unit values for multifamily buildings within each geographic location of the at least one geographic region.

13 . A method for predicting multi-family property rental income strength across geographic areas, the method comprising:

consolidating, by one or more processors, a plurality of location feature data sets comprising, for each respective geographic region of a plurality of geographic regions, a respective location feature data set relevant to the respective geographic region, the respective location feature data set comprising, for each location feature of a plurality of location features, a respective one or more location feature data values, wherein

the plurality of geographic regions is defined at multiple levels of granularity, and

location feature data values corresponding to different location features of the plurality of location features are available at different levels of granularity of the multiple levels of granularity;

storing the plurality of location feature data sets to a non-transitory computer readable medium;

transforming, by the one or more processors, a plurality of location feature data values of a set of location features of the plurality of location features from an original granularity level of the multiple levels of granularity as consolidated in the plurality of location feature data sets to another granularity level of the multiple levels of granularity;

storing, to the non-transitory computer readable medium, the transformed plurality of feature data values as part of the plurality of location feature data sets, wherein

the transforming populates the plurality of feature data sets with feature data values for each of the set of location features at each granularity level of the multiple levels of granularity:

identifying a plurality of output variables comprising, for each respective geographic region of at least a portion of the plurality of geographic regions, a respective output variable representing rental income per unit time in the respective geographic region;

by the one or more processors and using the plurality of location feature data sets and the plurality of output variables, training, for each respective geographic area of a plurality of geographic areas, at least one respective machine learning model of a plurality of machine learning models to predict i) an average rental income per unit value for multifamily buildings within the respective geographic area, and ii) a set of feature-level scores indicating a contribution strength of each location feature of at least a portion of the plurality of location features to the average rental income per unit value, wherein

the plurality of machine learning models comprise at least one machine learning model for each granularity level of the multiple levels of granularity;

in real time responsive to receiving input via a user interface screen presented to a user of a remote computing device, determining, by the one or more processors, a location score calculation data set representing a) at least one geographic region of the plurality of geographic regions and b) at least a portion of the plurality of location features;

applying, by the one or more processors, the location score calculation data set to one or more machine learning models of the plurality of machine learning models to obtain

a respective average rental income per unit value for multifamily buildings within each geographic region of the at least one geographic region, and

a respective set of feature-level scores within each geographic region of the at least one geographic region, wherein

the one or more machine learning models are relevant to the at least one geographic region,

for each geographic region of the at least one geographic region, rescaling, by the one or more processors, each feature-level score of the respective set of feature-level scores using a geographic region quality score scale to produce a set of scaled feature-level scores quantifying, for each location feature of the portion of the plurality of location features, a deviation of each scaled feature-level score from a corresponding geographic area encompassing the at least one geographic region; and

by the one or more processors in real-time responsive to receiving the input via the user interface screen, returning, for presentation at the user interface screen of the remote computing device,

at least one rental income prediction based on the respective average rental income per unit value for multifamily buildings within each geographic region of the at least one geographic region, and

feature score information corresponding to at least a portion of features of the respective set of scaled feature-level scores corresponding to one or more geographic locations regions of the at least one geographic region.

14 . The method of claim 13 , wherein the average rental income per unit value is calculated as a rental income strength on a log scale.

15 . The method of claim 14 , further comprising transforming, by the one or more processors, the rental income strength to a linear scale as a normalized location quality score.

16 . The method of claim 13 , wherein transforming the plurality of feature data values comprises, for each respective location feature of a subset of the plurality of location features, transforming the respective one or more location feature data values of each respective location feature data set relevant to at least two overlapping geographic regions of the plurality of geographic regions overlapping with a geographic area to produce a corresponding one or more location feature data values applicable to the geographic area, wherein the geographic area is at a different level of granularity of the multiple levels of granularity than the at least two overlapping geographic regions.

17 . The method of claim 13 , wherein returning the feature score information comprises returning, for each respective feature of the portion of features, a respective current metric for the respective feature at each geographic location of the at least one geographic region, and a respective score for the respective feature at each geographic location of the at least one geographic region.

18 . The method of claim 13 , further comprising, by the one or more processors:

identifying a plurality of missing feature data values of one or more location features of the plurality of location features corresponding to a first geographic area at a first granularity level of the multiple levels of granularity; and

filling in each missing feature data value of the plurality of missing feature data values.

19 . The method of claim 18 , wherein filling in each missing feature data value comprises, for at least a portion of the plurality of missing feature data values:

for each respective location feature of the one or more location features and for each respective feature data value of the portion of the plurality of missing feature data values corresponding to the respective location feature, impute an imputed feature value using one or more feature data values of the respective location feature corresponding to a second geographic area at the first granularity level, wherein the second geographic area is physically adjacent to the first geographic area.

20 . The method of claim 13 , further comprising:

by the one or more processors, storing, to a non-transitory computer-readable data store, the respective average rental income per unit value for multifamily buildings within each geographic location of the at least one geographic region;

wherein the at least one rental income prediction comprises a trended location score generated using past average rental income per unit values for multifamily buildings within each geographic location of the at least one geographic region.