OWL Check on Timeseries Event Sequence
I see one use-case where the customer is looking owl to detect for an event ID if Message (A, B, C) come in this pattern A --> B-- > C (80% of the time) at different times in a day and one day if it comes to B–>A–>C the owl should detect this?
is this possible? Event ID will be the same
Just to add A, B and C are in a single column
bineeshbabu
Posted 5 years ago · Edited 1 year ago·Last reply 5 years ago
8 comments
bineeshbabu
OP5 years ago · EditedThanks @kirk.haslbeck.collibra.com
This was really helpful
kirkhaslbeck
·5 years ago · EditedIf you use the DataFrame API you can write whatever spark code you like. If you want to use the Rule Builder you could write SQL like the above.
bineeshbabu
OP5 years ago · EditedThanks @kirk.haslbeck.collibra.com
Can we write such spark functions in owl rule tab or idea is to create a View and then write a SQL
kirkhaslbeck
·5 years ago · EditedIn the ML world this would be similar to a Markov Chain where you predict the sequence of events. While Collibra DQ does support time series outliers it does not support time series event sequences, at least not as an out of the box 1 click feature. One thing to consider for this unique case is to write a DQ rule that transposes by eventID and pivots on Message, similar to a GROUP by on eventId and then write a simple sql conditional check like below.
bineeshbabu
OP5 years ago · EditedThanks @kirk.haslbeck.collibra.com, Message A,B,C are not in single cell. They are in different row . something like below
kirkhaslbeck
·5 years ago · EditedIf A,B,C are in a single column and the column looks like this
event_id
a,b,c
a,b,c
d,e,f
a,b,c
b,c,a -> alert
then all you have to do is click OUTLIER -> Categorical and select event_id. The b,c,a would be a rare categorical, low frequency outlier.
bineeshbabu
OP5 years ago · EditedThis via some ETL process into a Database
BigDataBear
·5 years ago · EditedHi Bineesh, is this a Kafka Streaming question, or how are the “messages” coming through to DQ?