Temporal discrete event analytics system

1. A method for evaluating a computer system using temporal discrete event analytics, comprising:

generating sentences of discrete event sequences for a plurality of sensors;

building a sensor relationship network in response to generating the sentences of discrete event sequences;

analyzing the sensor relationship network to determine relationships between the plurality of sensors;

performing fault diagnosis based on the sensor relationship network and the relationships between the plurality of sensors; and

performing a remedial action to address the fault diagnosis;

wherein building the sensor relationship network further comprises:

generating sensor pairs from the plurality of sensors; and

applying a neural machine translation (NMT) model to quantify strength of invariant relationship between at least one source sensor and at least one target sensor, and

wherein applying the NMT model to quantify strength of invariant relationship further comprises:

comparing at least one translated sentence of the at least one source sensor and at least one sentence of the at least one target sensor; and calculating an invariant relationship score to quantify the strength a pairwise relationship of the at least one source sensor and the at least one target sensor.

2. The method as recited in claim 1 , further comprising:

receiving a corpus of multivariate discrete event sequences; and

extracting the discrete event sequences from the corpus of multivariate discrete event sequences.

3. The method as recited in claim 2 , further comprising:

performing first-order reduction of a number of sensor pairs in the plurality of sensors.

4. The method as recited in claim 1 , wherein generating sentences of discrete event sequences further comprises:

converting discrete event sequences to words; and converting the words to the sentences of discrete event sequences.

5. The method as recited in claim 4 , wherein converting the discrete event sequences to the words further comprises:

forming a symbolic word vocabulary.

6. The method as recited in claim 1 , further comprising:

generating word embeddings using a symbolic word vocabulary; and performing translation of the discrete event sequences on a sentence by sentence basis.

7. The method as recited in claim 1 , wherein the invariant relationship score is a bilingual evaluation understudy (BLEU) score.

8. The method as recited in claim 1 , wherein analyzing the sensor relationship network to determine relationships between the plurality of sensors further comprises:

identifying popular sensors that act as target sensors; and discarding the popular sensors.

9. The method as recited in claim 1 , wherein analyzing the sensor relationship network to determine relationships between the plurality of sensors further comprises:

determining at least one cluster of sensors; and determining whether the at least one cluster is connected with at least one particular sensor.

10. The method as recited in claim 1 , wherein building the sensor relationship network using the sentences of discrete event sequences further comprises:

building the sensor relationship network without domain knowledge.

11. A computer system for evaluating another computer system using temporal discrete event analytics, comprising:

a processor device operatively coupled to a memory device, the processor device being configured to:

generate sentences of discrete event sequences for a plurality of sensors;

build a sensor relationship network in response to generating the sentences of discrete event sequences;

analyze the sensor relationship network to determine relationships between the plurality of sensors;

perform fault diagnosis based on the sensor relationship network; and

perform a remedial action to address the fault diagnosis;

wherein building the sensor relationship network further comprises the processor device configured to:

generating sensor pairs from the plurality of sensors; and

applying a neural machine translation (NMT) model to quantify strength of invariant relationship between at least one source sensor and at least one target sensor, and

wherein applying the NMT model to quantify strength of invariant relationship further comprises the processor device configured to:

comparing at least one translated sentence of the at least one source sensor and at least one sentence of the at least one target sensor; and calculating an invariant relationship score to quantify the strength a pairwise relationship of the at least one source sensor and the at least one target sensor.

12. The system as recited in claim 11 , wherein the processor device is further configured to:

receive a corpus of multivariate discrete event sequences; and

extract the discrete event sequences from the corpus of multivariate discrete event sequences.

13. The system as recited in claim 11 , wherein the processor device is further configured to:

perform first-order reduction of a number of sensor pairs in the plurality of sensors.

14. The system as recited in claim 11 , wherein, when generating sentences of discrete event sequences, the processor device is further configured to:

convert discrete event sequences to words; and convert the words to the sentences of discrete event sequences.

15. The system as recited in claim 14 , wherein, when converting the discrete event sequences to the words, the processor device is further configured to:

form a symbolic word vocabulary.

16. The system as recited in claim 11 , wherein the processor device is further configured to:

generate word embeddings using a symbolic word vocabulary; and perform translation of the discrete event sequences on a sentence by sentence basis.