Artificial intelligence-based block embedding
A computer system and associated processes for grouping similar real estate properties into contiguous neighborhoods and generating neighborhood-specific models capable of estimating property values within their neighborhoods. An artificial intelligence system directed to using a graph neural network framework to identify relationships between different parcel groups based on similar property features and embed the parcel groups into low dimensional space vectors. The method can include generating a graph and features relevant to the parcel groups that can train an embedding function that generate an embedding vector for each parcel group in a geographic unit grouping, such as a census tract. Embedding vectors of two or more parcel groups can then be compared to each other to determine whether the parcel groups are similar or to determine a housing valuation of a parcel group.
1 . A computer-implemented method for embedding parcel groups, the computer-implemented method comprising:
obtaining a geographic unit grouping, wherein the geographic unit grouping comprises at least two parcel groups including a first parcel group and a second parcel group, wherein each of the first parcel group and the second parcel group comprises at least one parcel;
obtaining property-level data for the first and second parcel groups in the geographic unit grouping;
generating a graph model using the property-level data, wherein the graph model comprises one or more polygons for each of the first and second parcel groups that indicate a relationship between a geographic area of the first parcel group and a geographic area of the second parcel group;
generating property features for each of the first and second parcel groups using the property-level data;
training an embedding function to generate an embedding vector that include a number of dimensions that is less than a number of the at least two parcel groups based on inputting the property features and the one or more polygons of the graph model into a neighbor predicting artificial intelligence model, wherein the neighbor predicting artificial intelligence model is trained to determine a relationship between the at least two parcel groups, the embedding vector comprising a unique identification of a parcel group based on the relationship between the at least two parcel groups;
generating a first embedding vector of the first parcel group and a second embedding vector for the second parcel group using the trained embedding function;
applying the first and second embedding vectors as an input to an artificial intelligence model, wherein application of the first and second embedding vectors as the input to the artificial intelligence model causes the artificial intelligence model to produce an output; and
generating an outcome for the first parcel group based on the output.
2 . The method of claim 1 , wherein the property-level data comprises property data and census data.
3 . The method of claim 1 , wherein generating the property features for each of the first and second parcel groups comprises an aggregation of parcel features for each of the at least one parcels in the first and second parcel groups respectively.
4 . The method of claim 1 , further comprising generating a visualization of the first and second embedding vectors.
5 . The method of claim 1 , wherein the artificial intelligence model comprises one of an automated valuation model, a rental valuation model, or a neighborhood recommendation model.
6 . The method of claim 1 , wherein the outcome comprises one of a housing value, a rental value, or a recommendation for the first parcel group.
7 . The method of claim 1 , wherein training the embedding function comprises:
obtaining the neighbor predicting artificial intelligence model;
generating the embedding function using the property features and the one or more polygons of the graph model as an input to train the neighbor predicting artificial intelligence model; and
training the embedding function using the neighbor predicting intelligence model and the property features to generate the trained embedding function;
wherein generating the first embedding vector using the trained embedding function comprises:
determining a neighbor score between the first and second parcel groups corresponding to the property features using the neighbor predicting artificial intelligence model; and
inputting into the trained embedding function the property features of the first parcel group and, if the neighbor score meets a neighbor threshold, further inputting into the trained embedding function the property features of the second parcel group.
8 . The method of claim 7 , wherein determining the neighbor score further comprises weighing the property features unequally.
9 . The method of claim 1 , wherein obtaining the geographic unit grouping further comprises encoding each of the first and second parcel groups.
10 . The method of claim 1 , wherein the embedding vector comprise a dimensionally reduced vector of the first and second parcel groups, and wherein a dimension is encoded with the property feature of the first and second parcel group.
11 . The method of claim 10 , wherein the number of dimensions for each of the embedding vector comprises 32-dimensions.
12 . The method of claim 1 , further comprising:
generating a plurality of embedding vectors from the at least two parcel groups using the trained embedding function; and
training the artificial intelligence model using the plurality of embedding vectors prior to generating the first embedding vector and the second embedding vector.
13 . A system for parcel group embedding, the system comprising:
memory that stores computer-executable instructions; and
a processor in communication with the memory, wherein the computer-executable instructions, when executed by the processor, cause the processor to:
obtain a geographic unit grouping, wherein the geographic unit grouping comprises at least two parcel groups including a first parcel group and a second parcel group, wherein each of the first parcel group and the second parcel group comprises at least one parcel;
obtain property-level data for the first and second parcel groups in the geographic unit grouping;
generate a graph model using the property-level data, wherein the graph model comprises one or more polygons for each of the first and second parcel groups that indicate a relationship between a geographic area of the first parcel group and a geographic area of the second parcel group;
generate property features for each of the first and second parcel groups using the property-level data;
train an embedding function to generate an embedding vector that include a number of dimensions that are less than a number of the at least two parcel groups based on inputting the property features and the one or more polygons of the graph model into a neighbor predicting artificial intelligence model, wherein the neighbor predicting artificial intelligence model is trained to determine a relationship between the at least two parcel groups, the embedding vector comprising a unique identification of a parcel group based on the relationship between the at least two parcel groups;
generate a first embedding vector for the first parcel group and a second embedding vector for the second parcel group using the trained embedding function;
apply the first and second embedding vectors as an input to an artificial intelligence model, wherein application of the first and second embedding vectors as the input to the artificial intelligence model causes the artificial intelligence model to produce an output; and
generate an outcome for the first parcel group based on the output.
14 . The system of claim 13 , wherein the property-level data comprises property data and census data.
15 . The system of claim 13 , wherein the property features for each of the first and second parcel groups comprises an aggregation of parcel features for each of the at least one parcels in the first and second parcel groups respectively.
16 . The system of claim 13 , wherein the computer-executable instructions, when executed, further cause the processor to generate a visualization of the first and second embedding vectors.
17 . The system of claim 13 , wherein the artificial intelligence model comprises one of an automated valuation model, a rental valuation model, or a neighborhood recommendation model.
18 . The system of claim 13 , wherein the outcome comprises one of a housing value, a rental value, or a recommendation for the first parcel group.
19 . The system of claim 13 , wherein the computer-executable instructions, when executed, further cause the processor to train the embedding function by:
obtaining the neighbor predicting artificial intelligence model;
generating the embedding function using the property features and the one or more polygons of the graph model as an input to train the neighbor predicting artificial intelligence model; and
training the embedding function using the neighbor predicting intelligence model and the property features to generate the trained embedding function;
wherein the computer-executable instructions, when executed, further cause the processor to generate the first embedding vector using the trained embedding function by:
determining a neighbor score between the first and second parcel groups corresponding to the property features using the neighbor predicting artificial intelligence model; and
inputting into the trained embedding function the property features of the first parcel group and, if the neighbor score meets a neighbor threshold, further inputting into the trained embedding function the property features of the second parcel group.
20 . The system of claim 13 , wherein the computer-executable instructions, when executed, further cause the processor to obtain an encoded identification of the at least one parcels.
21 . The system of claim 13 , wherein the embedding vector comprise a dimensionally reduced vector of the first and second parcel groups, and wherein a dimension is encoded with the property feature of the first and second parcel group.
22 . The system of claim 21 , wherein each of the embedding vector comprises 32-dimensions.
23 . A non-transitory, computer-readable medium comprising computer-executable instructions for embedding parcel groups, wherein the computer-executable instructions, when executed by a computer system, cause the computer system to:
obtain a geographic unit grouping, wherein the geographic unit grouping comprises at least two parcel groups including a first parcel group and a second parcel group, wherein each of the first parcel group and the second parcel group comprises at least one parcel;
obtain property-level data for the first and second parcel groups in the geographic unit grouping;
generate a graph model using the property-level data, wherein the graph model comprises one or more polygons for each of the first and second parcel groups that indicate a relationship between a geographic area of the first parcel group and a geographic area of the second parcel group;
generate property features for each of the first and second parcel groups using the property-level data;
train an embedding function to generate an embedding vector that include a number of dimensions that are less than a number of the at least two parcel groups based on inputting the property features and the one or more polygons of the graph model into a neighbor predicting artificial intelligence model, wherein the neighbor predicting artificial intelligence model is trained to determine a relationship between the at least two parcel groups, the embedding vector comprising a unique identification of a parcel group based on the relationship between the at least two parcel groups;
generate a first embedding vector of the first parcel group and a second embedding vector for the second parcel group using the trained embedding function;
apply the first and second embedding vectors as an input to an artificial intelligence model, wherein application of the first and second embedding vectors as the input to the artificial intelligence model causes the artificial intelligence model to produce an output; and
generate an outcome for the first parcel group based on the output.