Systems and methods for classifying data in building automation systems

• US 9,411,327 B2
• Filed: 08/27/2012
• Issued: 08/09/2016
• Est. Priority Date: 08/27/2012
• Status: Active Grant

• Current Assignee: Tyco Fire & Security GmbH (Johnson Controls International plc)
• Original Assignee: Johnson Controls Technology Company (Johnson Controls International plc)
• Inventors: Park, Youngchoon
• Primary Examiner(s): Gami, Tejal
• Application Number:
  US13/595,945
• Publication Number:
  US 20140058539A1
• Time in Patent Office: 1,443 Days
• Field of Search: 700/50, 707/769, 707/741, 709/206
• US Class Current: 1/1
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 15/02 electric
• CPC Class Codes: G05B 2219/21088 Define name and address of I-O
• CPC Class Codes: G05B 2219/25011 Domotique, I-O bus, home automation, building automation
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A computerized method of assigning a building automation system point type to a plurality of unclassified data points is provided. The method includes receiving unclassified data points and attributes for each data point. The method includes receiving classifications for a first subset of the unclassified data points. Each classification associates a data point with a building automation system point type. The method includes generating a term set containing substrings that appear in the attributes. The method includes generating a first matrix describing a frequency that the substrings appear in the attributes. The method includes calculating an indicator of a probability that the presence of the selected substring results in the data point belonging to the selected point type. The method includes assigning a point type to a second subset by finding the substring and potential point type pair having the greatest indication of probability.

Claims

1. A computerized method of assigning a building automation system point type to a plurality of unclassified data points of a building automation system, the method comprising:

receiving, at a processing circuit, the plurality of unclassified data points from one or more subsystems or devices of the building automation system, each of the unclassified data points comprising at least one attribute containing a substring;

classifying each data point of a first subset of the unclassified data points by associating a building automation system point type of a plurality of building automation system point types with each data point of the first subset, wherein remaining unclassified data points form a second subset;

generating a term set containing the substrings that appear in the attributes of the first subset of data points;

calculating, for each substring in the term set, a frequency value indicating a frequency that the substring appears in the attributes of the first subset of data points associated with one of the building automation system point types;

repeating the calculating for each of the building automation system point types such that a frequency value is calculated for each unique combination of a selected substring and a selected building automation system point type;

generating a first matrix, each element of the first matrix corresponding to one of the unique combinations and containing the frequency value calculated for the unique combination;

using the elements of the first matrix to calculate, for each of the unique combinations, an indicator of a probability that the presence of the selected substring in the at least one attribute of a data point results in the data point belonging to the selected building automation system point type;

assigning a building automation system point type to each data point of the second subset by finding the substring and potential building automation system point type pair having the greatest indication of probability; and

using the data points of the second subset with the assigned building automation system point types to control one or more subsystems or devices of the building automation system.

2. The method of claim 1, wherein generating the first matrix comprises:

collecting data points of the first subset classified as the same building automation system point type;

concatenating the attributes of the data points of the first subset classified as the same building automation system point type;

calculating, for each of the substrings, the frequency the selected substring appears in the concatenated attributes associated with each building automation system point type; and

constructing the first matrix with columns of the concatenated attributes, rows of the substrings, and individual elements of the frequency that the selected substring appears the concatenated attributes associated with the selected building automation system point type.

3. The method of claim 2, wherein calculating the indicator of the probability comprises using a naive Bayes classifier.

4. The method of claim 3, wherein generating the first matrix comprises using latent semantic indexing.

5. The method of claim 2, wherein generating the first matrix further comprises weighting the frequency that the selected substring appears in the concatenated attribute associated with the selected building automation system point type using an inverse document frequency factor.

6. The method of claim 2, wherein the indicator of the probability, for each of the substrings and for each of the building automation system point types, is equal to the frequency that the selected substring appears in the concatenated attribute associated with the selected building automation system point type summed over the substrings and the concatenated attributes associated with each building automation system point type, divided by the number of the occurrences of any of the substrings in the concatenated attributes associated with the selected building automation system point type.

7. The method of claim 2, wherein generating the first matrix further comprises:

deconstructing the first matrix using singular value decomposition into the product of a second matrix, a third matrix, and a fourth matrix;

constructing a fifth matrix using portions of the second matrix, the third matrix, and the fourth matrix, wherein the portions are defined by a quality control constant, and wherein the fifth matrix is an approximation of the first matrix.

8. The method of claim 7, wherein the indicator of the probability, for each of the substrings and for each of the plurality of building automation system point types, is equal to the cosine distance of a vector describing the selected substring and a vector describing the selected building automation system point type.

9. The method of claim 1, further comprising:

receiving a second plurality of unclassified data points and at least one attribute for each data point;

generating at least one additional term set containing substrings that appear in the attributes of the at least one additional plurality of unclassified data points;

calculating a dissimilarity indicator using the term set and the at least one additional term set; and

generating a new term set containing substrings of the first subset of data points and the second plurality of unclassified data points when the dissimilarity indicator is greater than a threshold amount.

10. The method of claim 9, wherein the dissimilarity indicator comprises a Jaccard index.

11. A system for assigning a building automation system point type to a plurality of unclassified data points of a building automation system, the controller comprising:

a processing circuit configured to receive the plurality of unclassified data points from one or more subsystems or devices of the building automation system, each of the unclassified data points comprising at least one attribute containing a sub string;

wherein the processing circuit is further configured to classify each data point of a first subset of the unclassified data points by associating a building automation system point type of a plurality of building automation system point types with each data point of the first subset, wherein remaining unclassified data points form a second subset;

wherein the processing circuit is further configured to generate a term set containing the substrings that appear in the attributes of the first subset of data points;

wherein the processing circuit is further configured to calculate, for each substring in the term set, a frequency value indicating a frequency that the substring appears in the attributes of the first subset of data points associated with one of the building automation system point types;

wherein the processing circuit is further configured to repeat the calculation for each of the building automation system point types such that a frequency value is calculated for each unique combination of a selected substring and a selected building automation system point type;

wherein the processing circuit is further configured to generate a first matrix, each element of the first matrix corresponding to one of the unique combinations and containing the frequency value calculated for the unique combination;

wherein the processing circuit is further configured to use the elements of the first matrix to calculate, for each of the unique combinations, an indicator of a probability that the presence of the selected substring in the at least one attribute of a data point results in the data point belonging to the selected building automation system point type;

wherein the processing circuit is further configured to assign a building automation system point type to each data point of the second subset by finding the substring and potential building automation system point type pair having the greatest indication of probability; and

wherein the processing circuit is further configured to use the data points of the second subset with the assigned building automation system point types to control one or more subsystems or devices of the building automation system.

12. The system of claim 11, wherein generating the first matrix comprises:

collecting data points of the first subset classified as the same building automation system point type;

concatenating the attributes of the data points of the first subset classified as the same building automation system point type;

calculating, for each of the substrings, the frequency the selected substring appears in the concatenated attributes associated with each building automation system point type; and

constructing the first matrix with columns of the concatenated attributes, rows of the substrings, and individual elements of the frequency that the selected substring appears the concatenated attributes associated with the selected building automation system point type.

13. The system of claim 12, wherein calculating the indicator of the probability comprises using a naive Bayes classifier.

14. The system of claim 13, wherein generating the first matrix comprises using latent semantic indexing.

15. The system of claim 12, wherein generating the first matrix further comprises weighting the frequency that the selected substring appears in the concatenated attribute associated with the selected building automation system point type using an inverse document frequency factor.

16. The system of claim 12, wherein the indicator of the probability, for each of the substrings and for each of the building automation system point types, is equal to the frequency that the selected substring appears in the concatenated attribute associated with the selected building automation system point type summed over the substrings and the concatenated attributes associated with each building automation system point type, divided by the number of the occurrences of any of the substrings in the concatenated attributes associated with the selected building automation system point type.

17. The system of claim 12, wherein generating the first matrix further comprises:

deconstructing the first matrix using singular value decomposition into the product of a second matrix, a third matrix, and a fourth matrix;

constructing a fifth matrix using portions of the second matrix, the third matrix, and the fourth matrix, wherein the portions are defined by a quality control constant, and wherein the fifth matrix is an approximation of the first matrix.

18. The system of claim 17, wherein the indicator of the probability, for each of the substrings and for each of the plurality of building automation system point types, is equal to the cosine distance of a vector describing the selected substring and a vector describing the selected building automation system point type.

19. The system of claim 11, wherein the processing circuit is further configured to:

receive a second plurality of unclassified data points and at least one attribute for each data point;

generate at least one additional term set containing substrings that appear in the attributes of the at least one additional plurality of unclassified data points;

calculate a dissimilarity indicator using the term set and the at least one additional term set; and

generate a new term set containing substrings of the first subset of data points and the second plurality of unclassified data points when the dissimilarity indicator is greater than a threshold amount.

20. The system of claim 19, wherein the dissimilarity indicator comprises a Jaccard index.