Building risk analysis system with natural language processing for threat ingestion
A building management system includes one or more computer-readable storage media having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to receive threat events from one or more data sources, each threat event including a description and for each threat event determine whether the description for the threat event corresponds to one of a multiple predefined threat categories, generate a standardized threat object for the threat event using the corresponding predefined threat category, and in response to determining the description does not correspond to one of the predefined threat categories, process the description using a natural language processing engine to identify one of the predefined threat categories to be assigned to the threat event and generate a standardized threat object for the threat event using the predefined threat category identified by the natural language processing engine.
1. A building management system comprising:
one or more computer-readable storage media having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to:
receive threat events from one or more data sources indicating a potential threat to at least one of buildings, building equipment, people, or spaces within a building, each threat event comprising a description;
for each threat event:
determine whether the description for the threat event corresponds to one of a plurality of predefined threat categories;
in response to determining the description corresponds to one of the predefined threat categories, generate a standardized threat object for the threat event using the corresponding predefined threat category; and
in response to determining the description does not correspond to one of the predefined threat categories:
process the description using a natural language processing engine to identify one of the predefined threat categories to be assigned to the threat event; and
generate a standardized threat object for the threat event using the predefined threat category identified by the natural language processing engine.
2. The building management system of claim 1 , wherein determining whether the description for the threat event corresponds to one of the predefined threat categories comprises determining whether a stored data identifies a direct mapping of at least a portion of the description to one of the predefined threat categories.
3. The building management system of claim 1 , wherein the threat events are received from a plurality of data sources having a plurality of different data formats.
4. The building management system of claim 1 , wherein the instructions cause the one or more processors to generate a classifier for the natural language processing engine using historical threat data comprising a plurality of historical threat events each having a description, wherein the instructions cause the one or more processors to process the description using the classifier of the natural language processing engine.
5. The building management system of claim 4 , wherein the instructions cause the one or more processors to pre-process the historical threat data, pre-processing the historical threat data comprising at least one of the following:
filtering historical threat events from the set having a description longer than a first threshold length; or
filtering historical threat events from the set having a description shorter than a second threshold length.
6. The building management system of claim 5 , wherein the instructions cause the one or more processors to generate vector representations from the historical threat data subsequent to applying the labels.
7. The building management system of claim 5 , wherein the instructions cause the one or more processors to separate the historical threat data into a training data set and a test data set, wherein the instructions cause the one or more processors to train the classifier using the training data set and to test an accuracy of the classifier using the test data set.
8. The building management system of claim 5 , wherein the instructions cause the one or more processors to apply labels to the pre-processed historical threat data from among a set of predetermined labels, the set of predetermined labels associated with the predefined threat categories.
9. The building management system of claim 8 , wherein the instructions cause the one or more processors to apply the labels based on input from one or more users, the one or more processors configured to generate a user interface via which the input is received, wherein, for each historical threat event, the instructions cause the one or more processors to generate a suggested subset of the set of predetermined labels from which the one or more users can select the label to be applies to the historical threat event, wherein the instructions cause the one or more processors to generate the suggested subset of predetermined labels by performing a similarity analysis between the description of the historical threat event and the labels of the set of predetermined labels and including one or more predetermined labels from the set of predetermined labels having a highest similarity with the description of the historical threat event in the suggested subset.
10. The building management system of claim 9 , wherein performing the similarity analysis comprises performing a vector space similarity comparison between the description of the historical threat event and the set of predetermined labels and generating a similarity score for each predetermined label of the set of predetermined labels, wherein the instructions cause the one or more processors to select a one or more of the predetermined labels having the highest similarity scores for inclusion in the suggested subset.
11. A method for a risk analysis, the method comprising:
receiving threat events from one or more data sources indicating a potential threat to at least one of buildings, building equipment, people, or spaces within a building, each threat event comprising a description;
for each threat event:
determining whether the description for the threat event corresponds to one of a plurality of predefined threat categories;
in response to determining the description corresponds to one of the predefined threat categories, generating a standardized threat object for the threat event using the corresponding predefined threat category; and
in response to determining the description does not correspond to one of the predefined threat categories:
processing the description using a natural language processing engine to identify one of the predefined threat categories to be assigned to the threat event; and
generating a standardized threat object for the threat event using the predefined threat category identified by the natural language processing engine.
12. The method of claim 11 , wherein determining whether the description for the threat event corresponds to one of the predefined threat categories comprises determining whether a stored data identifies a direct mapping of at least a portion of the description to one of the predefined threat categories.
13. The method of claim 11 , wherein the threat events are received from a plurality of data sources having a plurality of different data formats.
14. The method of claim 11 , further comprising generating a classifier for the natural language processing engine using historical threat data comprising a plurality of historical threat events each having a description, wherein processing the description comprises using the classifier of the natural language processing engine.
15. The method of claim 14 , further comprising pre-processing the historical threat data, wherein pre-processing the historical threat data comprises at least one of the following:
filtering historical threat events from the set having a description longer than a first threshold length; or
filtering historical threat events from the set having a description shorter than a second threshold length.
16. The method of claim 15 , further comprising generating vector representations from the historical threat data subsequent to applying the labels.
17. The method of claim 15 , further comprising separating the historical threat data into a training data set and a test data set, wherein the method further comprises training the classifier using the training data set and to test an accuracy of the classifier using the test data set.
18. The method of claim 15 , further comprising applying labels to the pre-processed historical threat data from among a set of predetermined labels, the set of predetermined labels associated with the predefined threat categories.
19. The method of claim 18 , wherein applying the labels is based on input from one or more users, wherein the method further comprises generating a user interface via which the input is received, wherein, for each historical threat event, the method further comprises generating a suggested subset of the set of predetermined labels from which the one or more users can select the label to be applies to the historical threat event, wherein generating the suggested subset of predetermined labels comprises performing a similarity analysis between the description of the historical threat event and the labels of the set of predetermined labels and including one or more predetermined labels from the set of predetermined labels having a highest similarity with the description of the historical threat event in the suggested subset.
20. A building risk analytics system comprising:
one or more computer-readable storage media communicably coupled to one or more processors and configured to store instructions; and
the one or more processors configured to execute the instructions to:
receive threat events from one or more data sources indicating a potential threat to at least one of buildings, building equipment, people, or spaces within a building, each threat event comprising a description, wherein the threat events are received from a plurality of data sources having a plurality of different data formats;
for each threat event:
determine whether the description for the threat event corresponds to one of a plurality of predefined threat categories;
in response to determining the description corresponds to one of the predefined threat categories, generate a standardized threat object for the threat event using the corresponding predefined threat category; and
in response to determining the description does not correspond to one of the predefined threat categories:
process the description using a natural language processing engine to identify one of the predefined threat categories to be assigned to the threat event; and
generate a standardized threat object for the threat event using the predefined threat category identified by the natural language processing engine.