Multi-dimensional impact detect and diagnosis in cellular networks
View Patent ↗A method for multi-dimensional impact detection and diagnosis of networks. The methods and systems dynamically explore only those network feature combinations that are likely to have problems by using a summary structure Sketch, for example. The method and systems capture fine-grained anomalies at a coarse level which allows for on-demand zoom into the finer-grained levels for further analysis.
1. An apparatus comprising:
a processor; and
a memory coupled with the processor, the memory storing executable instructions that when executed by the processor cause the processor to effectuate operations comprising:
obtaining a record associated with a property;
determining the record matches a feature of a first node, wherein the first node is in a lattice that comprises a set of nodes that are organized based on feature values of the property;
updating a first statistic of the first node to a second statistic of the first node;
determining that the difference between the first statistic and the second statistic has reached a threshold;
based on the difference between the first statistic and the second statistic reaching the threshold, adding a child node to the first node;
determining that the child node is anomalous;
based on determining that the child node is anomalous, determining root cause features, wherein the root cause features are a subset of the records; and
transmitting a message comprising the root cause of the anomaly for the first node.
2. The apparatus of claim 1 , wherein the determining the child node is anomalous is based on determining whether any dimension in the respective sketch exceeds kσ, wherein σ is standard deviation, and wherein k is a constant.
3. The apparatus of claim 1 , the operations further comprising initializing a new lattice node using values of the first node based on an independence assumption, wherein the first node is a parent node.
4. The apparatus of claim 1 , the operations further comprising restricting connections to network devices based on the determined root cause features.
5. The apparatus of claim 1 , wherein the determining the root cause features is based on a compressive sensing method.
6. The apparatus of claim 1 , wherein the determining the root cause features is based on a compressive sensing method, and wherein the compressing sensing method comprises input that comprises enumeration of possible feature vectors and their associated performance statistics.
7. The apparatus of claim 1 , wherein the determining the root features is based on a neural network method.
8. A non-transitory computer readable storage medium storing computer executable instructions that when executed by a computing device cause said computing device to effectuate operations comprising:
obtaining a record associated with a property;
determining the record matches a feature of a first node, wherein the first node is in a lattice that comprises a set of nodes that are organized based on feature values of the property;
updating a first statistic of the first node to a second statistic of the first node;
determining that the difference between the first statistic and the second statistic has reached a threshold;
based on the difference between the first statistic and the second statistic reaching the threshold, adding a child node to the first node;
determining that the child node is anomalous;
based on determining that the child node is anomalous, determining root cause features, wherein the root cause features are a subset of the records; and
transmitting a message comprising the root cause of the anomaly for the first node.
9. The non-transitory computer readable storage medium of claim 8 , wherein the determining the child node is anomalous is based on determining whether any dimension in the respective sketch exceeds kσ, wherein σ is standard deviation, and wherein k is a constant.
10. The non-transitory computer readable storage medium of claim 8 , the operations further comprising initializing a new lattice node using values of the first node based on an independence assumption, wherein the first node is a parent node.
11. The non-transitory computer readable storage medium of claim 8 , the operations further comprising restricting connections to network devices based on the determined root cause features.
12. The non-transitory computer readable storage medium of claim 8 , wherein the determining the root cause features is based on a compressive sensing method.
13. The non-transitory computer readable storage medium of claim 8 , wherein the determining the root cause features is based on a compressive sensing method, and wherein the compressing sensing method comprises input that comprises enumeration of possible feature vectors and their associated performance statistics.
14. The non-transitory computer readable storage medium of claim 8 , wherein the determining the root features is based on a neural network method.
15. A method comprising:
obtaining a record associated with a property;
determining the record matches a feature of a first node, wherein the first node is in a lattice that comprises a set of nodes that are organized based on feature values of the property;
updating a first statistic of the first node to a second statistic of the first node;
determining that the difference between the first statistic and the second statistic has reached a threshold;
based on the difference between the first statistic and the second statistic reaching the threshold, adding a child node to the first node;
determining that the child node is anomalous;
based on determining that the child node is anomalous, determining root cause features, wherein the root cause features are a subset of the records; and
transmitting a message comprising the root cause of the anomaly for the first node.
16. The method of claim 15 , wherein the determining the child node is anomalous is based on determining whether any dimension in the respective sketch exceeds kσ, wherein σ is standard deviation, and wherein k is a constant.
17. The method of claim 15 , the operations further comprising restricting connections to network devices based on the determined root cause features.
18. The method of claim 15 , wherein the determining the root cause features is based on a compressive sensing method.
19. The method of claim 15 , wherein the determining the root cause features is based on a compressive sensing method, and wherein the compressing sensing method comprises input that comprises enumeration of possible feature vectors and their associated performance statistics.
20. The method of claim 15 , wherein the determining the root features is based on a neural network method.