Computer processes for clustering properties into neighborhoods and generating neighborhood-specific models
A computer system and associated processes are disclosed for grouping similar real estate properties into contiguous neighborhoods, and for generating neighborhood-specific models capable of estimating property values within their respective neighborhoods. A clustering component uses various sources of property-level data to group properties based on measures of property similarity. For example, the clustering component may use features extracted from property images to identify properties with similar characteristics. As another example, the clustering component may measure property similarity based on how frequently specific properties are designated as comparable in appraisal reports. A model generator uses a machine learning process to determine, for specific neighborhoods, correlations between property attributes and values, and uses these correlations to generate the neighborhood specific models.
1. A computing system comprising one or more computing devices programmed with executable code to implement at least:
a feature extraction component that extracts property features from property-level data associated with real estate properties; and
a clustering component that uses at least the extracted property features to group properties into neighborhoods based on property similarity such that each neighborhood is a contiguous geographic neighborhood, wherein the clustering component gives different amounts of weight to different property features for purposes of measuring property similarity;
wherein the property-level data comprises appraisal reports that identify specific properties as comparable, and the clustering component uses the identifications of comparable properties as a factor in measuring property similarity.
2. The computing system of claim 1 , wherein the feature extraction component analyzes property images to extract features associated with characteristics and conditions of houses, and the clustering component uses said features associated with characteristics and conditions of houses to measure property similarity.
3. The computing system of claim 1 , wherein the clustering component generates, for a property pair composed of a first property and a second property, an occurrence count value representing a number of times the first and second properties have been designated as comparable in the appraisal reports, and further uses said count value as a factor for measuring the property similarity of the first and second properties.
4. The computing system of claim 1 , wherein the clustering component is configured to measure similarity between two neighbor blocks, each of which comprises a plurality of properties, based at least in part on appraisal report based links between the properties in the respective neighbor blocks.
5. The computing system of claim 4 , wherein the clustering component is configured to merge neighbor blocks based at least in part on measures of similarity between the neighbor blocks.
6. The computing system of claim 4 , wherein the model generation component is configured to use a machine learning process to determine, for a specific neighborhood, correlations between property attributes and property values in the specific neighborhood, and is configured to generate weight values specifying amounts of weight to give to specific property attributes in estimating property values in the specific neighborhood.
7. The computing system of claim 1 , wherein the computing system is further programmed with the executable code to implement a model generation component that generates, for individual neighborhoods formed by the clustering component, neighborhood-specific Automated Valuation Models (AVMs), each neighborhood-specific AVM configured to estimate values of properties in its respective neighborhood.
8. A computing system comprising one or more computing devices programmed with executable program code to implement at process that comprises:
identifying a plurality of neighbor blocks within a geographic region, each neighbor block comprising a plurality of residential properties;
creating links between pairs of the neighbor blocks based on appraisal reports, each link based on a property in one of the neighbor blocks being identified in an appraisal report as comparable to another property in another neighbor block; and
applying a clustering algorithm to the neighbor blocks to form at least one contiguous geographic neighborhood, wherein the clustering algorithm groups together neighbor blocks based at least partly on numbers of said links formed between the neighbor blocks.
9. The computing system of claim 8 , wherein the clustering algorithm additionally takes into consideration data extracted from property images.
10. The computing system of claim 8 , wherein the process comprises generating a normalized score representing a strength of the links created between two of the neighbor blocks, wherein the clustering algorithm uses at least the normalized score to determine whether to group together the two neighbor blocks into a common neighborhood.
11. The computing system of claim 8 , wherein applying the clustering algorithm comprises determining whether to merge together two neighbor blocks to form a new neighbor block.
12. The computing system of claim 11 , further comprising, after a neighbor block merge operation, reapplying the clustering algorithm, and determining whether to merge any additional neighbor blocks.
13. The computing system of claim 12 , wherein determining whether to merge any additional neighbor blocks comprises, for a selected neighbor block formed by merging two or more other neighbor blocks, determining whether a strength of the links within the selected neighbor block exceeds a strength of links between the selected neighbor block and other neighbor blocks.
14. The computing system of claim 8 , wherein the process further comprises generating, for a neighborhood formed by application of the clustering algorithm, a neighborhood-specific model configured to estimate values of properties in the neighborhood.
15. The computing system of claim 14 , wherein generating the neighborhood-specific model comprises using a machine learning process to determine, for the neighborhood, correlations between property attributes and property values in the neighborhood, and generating weight values specifying amounts of weight to give to specific property attributes in estimating property values in the neighborhood.
16. A process performed by one or more computing devices under control of executable program instructions, the process comprising:
extracting property features from property-level data associated with real estate properties, said property-level data comprising appraisal reports that identify specific properties as comparable; and
grouping properties into neighborhoods based at least partly on property similarity such that each neighborhood is a contiguous geographic neighborhood, wherein grouping the properties into neighborhoods comprises using the identifications of comparable properties in the appraisal reports as a factor in measuring property similarity, and additionally comprises giving different amounts of weight to different ones of said extracted property features.
17. The process of claim 16 , wherein extracting said property features comprises extracting, from property images, features associated with characteristics and conditions of houses, and wherein grouping the properties into neighborhoods comprises using said features associated with characteristics and conditions of houses as a factor for measuring property similarity.
18. The process of claim 16 , wherein grouping the properties into neighborhoods comprises, for a property pair composed of a first property and a second property, generating an occurrence count value representing a number of times the first and second properties have been designated as comparable in the appraisal reports, and further comprises using said occurrence count value as a factor for measuring the property similarity of the first and second properties.
19. The process of claim 16 , wherein grouping the properties into neighborhoods comprises measuring similarity between two neighbor blocks, each of which comprises a plurality of properties, based at least in part on appraisal report based links between the properties in the respective neighbor blocks.
20. The process of claim 19 , wherein grouping the properties into neighborhoods further comprises merging neighbor blocks based at least in part on measures of similarity between the neighbor blocks.
21. The process of claim 16 , further comprising generating, for a first neighborhood of said neighborhoods, a neighborhood-specific Automated Valuation Model (AVM) configured to estimate values of properties in the first neighborhood.