System for clustering and aggregating data from multiple sources

• US 10,026,114 B2
• Filed: 01/12/2015
• Issued: 07/17/2018
• Est. Priority Date: 01/10/2014
• Status: Active Grant

• Current Assignee: Quest Analytics LLC
• Original Assignee: BetterDoctor, Inc. (Quest Analytics LLC)
• Inventors: Tolvanen, Tapio, Tulla, Ari, Hao, Tele
• Primary Examiner(s): Uddin, Mohammed R
• Application Number:
  US14/594,933
• Publication Number:
  US 20150199744A1
• Time in Patent Office: 1,282 Days
• Field of Search: 707737, 707726, 707748, 707999102
• CPC Class Codes: G06F 16/9537 Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records

Abstract

Systems and methods are provided for receiving, aggregating, and analyzing data to develop caregiver rankings, recommendations, and other information that care seekers may use to connect with caregivers for services, or for caregivers to use to connect with care seekers. Sample data can be obtained from a plurality of data sources, processed to form data clusters, aggregated to form data records, and provided to a care seeker searching for a caregiver or medical facility.

Claims

1. A method of aggregating entity data from a plurality of sources, the method comprising:

obtaining sample data from a plurality of data sources, the sample data corresponding to a plurality of entities that have not been previously identified, wherein samples from multiple data sources correspond to a same entity;

processing the samples to identify a plurality of fields corresponding to each sample;

identifying a first partial cluster of the samples based on a first set of rules that relies on a geographical indicator, the first partial cluster comprising a first subset of fields from a first sample for comparison with a second sample on the first subset of fields to determine whether the samples correspond with a same entity, wherein identifying the first partial cluster includes;

determining whether the second sample is in the first partial cluster by;

determining a first field distance between a first field of the first subset of fields of the first sample and a first field of the first subset of fields of the second sample;

calculating a first metric based on the first field distance; and

adding the second sample to the first metric when the first metric is within a first threshold;

identifying a second partial cluster of the samples based on a second set of rules that relies on a user identifier, the second partial cluster comprising a second subset of fields from the first sample for comparison with the second sample on the second subset of fields to determine whether the samples correspond with the same entity, wherein identifying the second partial cluster includes;

determining whether the second sample is in the second partial cluster by;

determining a second field distance between a second field of the second subset of fields of the first sample and a second field of the second subset of fields of the second sample;

calculating a second metric based on the second field distance; and

adding the second sample to the second metric when the second metric is within a second threshold;

initiating an aggregation process that determines that the second partial cluster of the samples corresponds to the same entity as the first partial cluster of the samples, wherein the aggregation process generates a full entity cluster that corresponds to the same entity having the user identifier for the first field and the geographical indicator for the second field; and

storing the user identifier and the geographical indicator of the full entity cluster into a first record of a database.

2. The method of claim 1, the method further comprising:

identifying which data source the samples are from;

determining confidence values corresponding to the data source; and

using the confidence values to determine the user identifier and the geographical indicator.

3. The method of claim 2, wherein at least one field of a sample from the data source is not used when the confidence values for the data source are below a confidence threshold.

4. The method of claim 2, further comprising:

receiving feedback from a first entity regarding the first record; and

computing the confidence values based on the feedback.

5. The method of claim 1, wherein the first threshold, the first field, and the first metric are specified by the first set of rules.

6. The method of claim 1, wherein the first set of rules specifies multiple fields for clustering the samples, each of the specified fields having a corresponding field distance.

7. The method of claim 6, wherein calculating the first metric based on the first field distance includes:

calculating a weighted average of the first field distance and one or more other field distances of the corresponding field distances, wherein weights of at least two field distances are different.

8. The method of claim 1, further comprising:

for each of one or more additional sets of rules;

identifying an additional cluster of the samples as corresponding to the first entity;

comparing the samples of the first cluster and the one or more additional clusters to determine samples in a first optimum cluster; and

determining the user identifier and the geographical indicator for the first entity by comparing the fields of samples in the first optimum cluster.

9. The method of claim 8, wherein at least two different sets of rules specify different fields to use for clustering the samples.

10. The method of claim 1, wherein the first field distance and the second field distance are determined using local sequence comparison that identifies similarities between corresponding fields in the first sample or the second sample corresponding with the first partial cluster or the second partial cluster.

11. The method of claim 1, further comprising:

determining first representative fields of the first partial cluster by comparing the fields of the samples in the first partial cluster; and

determining second representative fields of the second partial cluster by comparing the fields of the samples in the second partial cluster,

wherein the one or more field distances are determined using one or more corresponding fields of the first representative fields and the second representative fields.

12. The method of claim 1, further comprising:

determining that a first data source is associated with a higher confidence value than a second data source, the first data source and second data source included in the plurality of data sources;

determining that the first data source is an origin of data for the first record in the database based in part on the higher confidence value;

calculating a first decay rate for a first confidence value for the first data source and a second decay rate of a second confidence value for the second data source;

determining that the first data source is less accurate after a time than the second data source based in part on the first confidence value being less than the second confidence value; and

altering the first record in the database to correspond with the second data source in the database based on the first and second confidence values.

13. The method of claim 1, further comprising:

identifying the first record of the database to include first feedback;

comparing the first feedback associated with the first record of the database with second feedback associated with other records of the database;

generating a ranking of the records in the database based in part on the first feedback associated with the first record and the second feedback associated with the other records; and

enabling the ranking to display on a graphical user interface.

14. The method of claim 1, further comprising:

identifying the first record of the database, wherein the first record includes the first field from a first data source in the plurality of data sources;

obtaining new sample data from a second data source in the plurality of data sources, wherein the new sample data includes a different value for the first field compared to the first record;

determining a new confidence value associated with the new sample data, wherein an existing confidence value is associated with the first data source or the first field corresponding with the first record of the database;

determining that the new confidence value associated with the new sample data is less than the existing confidence value associated with the data source or first field corresponding with the first record of the database; and

maintaining the first record as unchanged despite the new sample data.

15. The method of claim 1, further comprising:

upon analyzing the full entity cluster, identifying a second entity cluster from the full entity cluster, wherein the second entity cluster corresponds with the full entity cluster by the geographical indicator for the full entity cluster; and

duplicating and storing the geographical indicator for the full entity cluster as a new geographical indicator for the second entity cluster.

16. The method of claim 15, wherein a second entity name field of the second entity cluster corresponds to a third field of the full entity cluster.

17. The method of claim 1, wherein other fields of the first sample and the second sample are telephone numbers.

18. The method of claim 1, wherein the aggregation process to generate the full entity cluster is performed in a hierarchy optimum cluster comprising:

determining the first field comprises the user identifier as a first most frequent entry with a first highest quality score in the first partial cluster; and

determining the second field comprises the geographical indicator as a second most frequent entry with a second highest quality score in the second partial cluster.

19. The method of claim 1, wherein the first partial cluster and second partial clusters are stored and reused in a different aggregation process to generate the full entity cluster.

20. The method of claim 1, wherein the aggregation process comprises:

determining a first frequency for each single entry of the geographical indicator from the first partial cluster;

generating a data object for the full entity cluster that includes the single entry of the geographical indicator when the first frequency for the single entry exceeds a third threshold;

determining a second frequency for each single entry of the user identifier or the name from the second partial cluster; and

updating the data object for the full entity cluster that includes the single entry of the user identifier or the name when the second frequency for the single entry exceeds a fourth threshold.

21. The method of claim 20, wherein the aggregation process implements a phonetic algorithm.

22. A computer product comprising a non-transitory computer readable medium embodying thereon a set of instructions, which when executed by a computer system cause the computer system to perform the steps of:

obtaining sample data from a plurality of data sources, the sample data corresponding to a plurality of entities that have not been previously identified, wherein samples from multiple data sources correspond to a same entity;

processing the samples to identify a plurality of fields corresponding to each sample;

identifying a first partial cluster of the samples based on a first set of rules that relies on a geographical indicator, the first partial cluster comprising a first subset of fields from a first sample for comparison with a second sample on the first subset of fields to determine whether the samples correspond with a same entity, wherein identifying the first partial cluster includes;

determining whether the second sample is in the first partial cluster by;

determining a first field distance between a first field of the first subset of fields of the first sample and a first field of the first subset of fields of the second sample;

calculating a first metric based on the first field distance; and

adding the second sample to the first metric when the first metric is within a first threshold;

identifying a second partial cluster of the samples based on a second set of rules that relies on a user identifier, the second partial cluster comprising a second subset of fields from the first sample for comparison with the second sample on the second subset of fields to determine whether the samples correspond with the same entity, wherein identifying the second partial cluster includes;

determining whether the second sample is in the second partial cluster by;

determining a second field distance between a second field of the second subset of fields of the first sample and a second field of the second subset of fields of the second sample;

calculating a second metric based on the second field distance; and

adding the second sample to the second metric when the second metric is within a second threshold;

initiating an aggregation process that determines that the second partial cluster of the samples corresponds to the same entity as the first partial cluster of the samples, wherein the aggregation process generates a full entity cluster that corresponds to the same entity having a user identifier for the first field and the geographical indicator for the second field; and

storing the user identifier and the geographical indicator of the full entity cluster into a first record of a database.

23. The computer product of claim 22, the steps further comprising:

identifying which data source a sample is from;

determining confidence values corresponding to the data source; and

using the confidence values to determine the user identifier and the geographical indicator for the first entity.

24. The computer product of claim 22, wherein at least one field of a sample from the data source is not used when the confidence values for the data source are below a confidence threshold.

25. The computer product of claim 22, further comprising:

receiving feedback from the first entity regarding the first record; and

computing the confidence values based on the feedback.