US 9846887

US Patent 9846887, titled "Discovering neighborhood clusters and uses therefor," was assigned to Carnegie Mellon University. The inventors are Justin Cranshaw, Raz Schwartz, Jason I. Hong, and Norman Sadeh-Koniecpol. The application was filed on August 30, 2013, and the patent was issued on December 19, 2017.

Abstract:
The patent describes computer-based systems and methods for identifying neighborhood clusters within a geographic region. These clusters are defined by a mix of venues and are determined using venue check-in data. The mix of venues can be based on the social similarity between pairs of venues, be representative of specific neighborhood typologies, or reflect temporal check-in pattern types, or a combination of these factors. The discovered neighborhood clusters, derived from venue check-in data, can then be utilized for various commercial and civic purposes.

Plain-Language Overview of Independent Claims:

Independent Claim 1: A System for Discovering Venue Clusters
This claim describes a computer-based system designed to identify geographic clusters of venues within a larger geographic area, using data about when people "check in" to venues. The system includes:

1. A computer database that stores venue check-in data from many different visitors for various venues in the region. This check-in data can come from mobile apps, point-of-sale transactions, venue ratings, or reviews.
2. One or more computer processors that are programmed to perform the following steps:
   • For each venue, create a "check-in intensity vector." This vector measures how often or intensely specific visitors (or groups of visitors) checked into that venue over a set period.
   • Create a "pairwise venue similarity matrix" for all the venues. This matrix assigns a similarity score to every pair of venues. This score is determined by both the geographical distance between the two venues and their "social distance."
   • Identify two or more distinct geographic clusters of venues using this similarity matrix. Each cluster will contain a mix of one or more venues. The "social distance" is specifically determined by whether the pair of venues is visited by common individuals or groups of visitors, and this similarity score might be zero if venues are too far apart or not among each other's closest neighbors.

Independent Claim 2: A Method for Discovering Venue Clusters
This claim outlines a computer-implemented method for achieving the same goal as the system in Claim 1 – identifying geographic clusters of venues in a region using venue check-in data. The method involves:

1. Storing venue check-in data from multiple visitors for various venues in a computer database.
2. Utilizing one or more processors to carry out these actions:
   • Generating a "check-in intensity vector" for each venue based on the stored check-in data, indicating the intensity of check-ins by visitors or groups over a specific time.
   • Generating a "pairwise venue similarity matrix" that contains similarity scores for each pair of venues. These scores are calculated based on both the geographical distance and the social distance between the venues.
   • Identifying two or more geographic clusters of venues in the region, using the generated similarity matrix. Each cluster comprises a mix of one or more venues. The social distance is determined by common visitors to the venues, and the similarity score can be zero if venues are geographically too distant.

Independent Claim 3: A System for Discovering Sub-Region Clusters
This claim describes a computer system for identifying geographic clusters, but instead of individual venues, it focuses on clustering sub-regions within a larger geographic area, where each sub-region itself contains multiple venues. The system includes:

1. A computer database that stores venue check-in data for venues located within these multiple sub-regions.
2. One or more computer processors that are programmed to:
   • Generate a "check-in intensity vector" for each sub-region. This vector reflects the cumulative number of times visitors checked into any venues within that sub-region over a period.
   • Generate a "pairwise similarity matrix" for these sub-regions. This matrix provides a similarity score for each pair of sub-regions, based on the likeness of their respective check-in intensity vectors.
   • Identify two or more geographic clusters of sub-regions based on this similarity matrix. Each resulting cluster is composed of a mix of one or more sub-regions. The grouping of sub-regions can be based on social similarity (i.e., common users checking into venues across those sub-regions), or whether the sub-regions represent certain geographic area types, or based on temporal check-in patterns.

No active dockets pertaining to US Patent 9846887 were found in the CAFC 2026 dockets.