...
The operator can save the learned tree into a database via the second output port. It continuously put out a serialized version of the tree which can be used to save it in a database, e.g. a MongoDB. When the analysis is started again, the information from the database can be used. This is useful, if a lot of knowledge is learned and if it would take a while to gain this knowledge again. On the startup, the information from the database is read a single time (if an information is available) and written into the operator. The PQL query in the codeblock shows, how to use the backup functionality.
| Code Block | ||
|---|---|---|
| ||
#PARSER PQL
#DEFINE BACKUPSCHEMA [['tree', 'String'],['backupId', 'String']]
#RUNQUERY
/// Backup-Daten aus der Datenbank auslesen
backupMongo = MONGODBSOURCE({
database = 'odysseus',
port = 27017,
host = 'localhost',
collectionname = 'condition'
}
)
/// Backup-Daten in Tupel konvertieren
backupTuple = KEYVALUETOTUPLE({
schema=${BACKUPSCHEMA},
TYPE = 'Backup',
KEEPINPUT = 'false'
},
backupMongo
)
stateAnalysis = RARESEQUENCE({
treedepth = 100,
minrelativefrequencyPath = 0.1,
firsttupleisroot = 'true',
UNIQUEBACKUPID = 'rareSequence_1'
},
state,
backupTuple
)
analysis = SELECT({PREDICATE = 'anomalyScore > 0'}, stateAnalysis)
/// Backup-Daten von Port 1 in ein Key-Value Objekt konvertieren
keyValueOp = TUPLETOKEYVALUE({
type='KEYVALUEOBJECT'
},
1:stateAnalysis
)
/// Backup-Daten in der Datenbank speichern
mongoSink = MONGODBSINK({
database = 'odysseus',
port = 27017,
host = 'localhost',
collectionname = 'condition',
batchsize = 1,
deletebeforeinsert = 'true',
deleteequalattribute = 'backupId'
},
keyValueOp
) |