logEvidence: Handle the relationship between evidence sets and student...

In ralmond/EABN: Evidence Accumulation Bayes Net Engine

Handle the relationship between evidence sets and student records.

Description

A StudentRecord differs from the baseline student record according to how many EvidenceSet objects have been incorporated into the estimate. These functions tie and student record and evidence set together.

Usage

logEvidence(eng, rec, evidMess)
seqno(x)
seqno(x) <- value
evidence(x)
evidence(x) <- value

Arguments

eng	A BNEngine which is currently not used (could later be used to save the evidence to a database).
rec	A StudentRecord into which the evidence will be incorporated.
evidMess	A EvidenceSet which will be associated with the student record.
x	An EvidenceSet object.
value	For seqno(x) <- value, an integer giving a new sequence number. For evidence(x) <- value, a character vector giving the sequence of evidence ID.

Details

There are several fields in the StudentRecord class which need to be updeated in the face of new evidence.

context and timestamp

These needs to be set to the values in the new evidence message.

seqno

This needs to be incremented.

evidence

The new evidence needs to be prepended to this list.

prev_id and "_id"

The prev_id needs to point to the old field and the "_id" is set to NA (it will be updated on save).

In the case of the BNEngineMongo, the IDs in question are the database ids for these objects so that they can be easily found. The function m_id For the BNEngineNDB case presumably some external system is issuing IDs to evidence sets and student records.

The evidence field of a StudentRecord is a list of IDs (m_id) for the accumulated evidence.

The seqno field is an optional ordering used to track the order in which evidence sets were incorporated into the student model. The value of seqno gives the number of evidence sets incorporated into the recrod.

The logEvidence function sets the sequence number of the evidence message to one more than the last sequence number for the student record. If no m_id exists for the record (no database mode), then one is generated by concatenating the uid and the seqno.

Value

The updateRecord returns a new StudentRecord object, which points back to the old one.

The logEvidence function returns the modified EvidenceSet.

The function seqno returns an integer (or NA if has not been set).

The function evidence returns a character vector giving the IDs (m_id) of the encorpated evidence sets.

Note

This is largely untested code for future fast retraction of evidence.

The prev_id field of the StudentRecord should leave a trace of previous student records in the database, including old serialized models. This should allow the scoring engine to quickly jump back in time.

The evidence field provides a list of the m_ids of all the incorporated evidence sets. This should enable one or more evidence sets to be replaced and the student model to be recalculated.

Author(s)

Russell Almond

See Also

BNEngine, EvidenceSet, EvidenceSet StudentRecord, handleEvidence P4Message

Examples

sess <- RNetica::NeticaSession()
RNetica::startSession(sess)
Nethouse <- PNetica::BNWarehouse(sess=sess)

recset <- StudentRecordSet(warehouse=Nethouse,db=MongoDB(noMongo=TRUE))

sr0 <-
    StudentRecord("S1","*baseline*",as.POSIXct("2020-03-30 09:00:00"))
seqno(sr0) <- 0
sr0 <- saveSR(recset,sr0) # Sets the m_id

e1 <- EvidenceSet(uid="S1",app="Test",context="PPcompEM",
                  obs=list("CompensatoryObs"="Right"))

e2 <- EvidenceSet(uid="S1",app="Test",context="PPdurAttEM",
                  obs=list("Attempts"=2,"Duration"=38.3))

stopifnot(is.na(seqno(e1)), seqno(sr0)==0L)
stopifnot(length(evidence(sr0))==0L)

e1 <- logEvidence(NULL,sr0,e1)
stopifnot(seqno(e1)==1L,!is.na(m_id(e1)))

sr1 <- updateRecord(sr0,e1)
stopifnot(is.na(m_id(sr1)),sr1@prev_id==m_id(sr0))
sr1 <- saveSR(recset,sr1)    # Sets the m_id

stopifnot(length(evidence(sr1))==1L,any(m_id(e1)==evidence(sr1)))
stopifnot(context(sr1)==context(e1),timestamp(sr1)==timestamp(e1))

e2 <- logEvidence(NULL,sr1,e2)
stopifnot(seqno(e2)==2L,!is.na(m_id(e2)))

sr2 <- updateRecord(sr1,e2)
stopifnot(is.na(m_id(sr2)),sr2@prev_id==m_id(sr1))
sr2 <- saveSR(recset,sr2) # Sets the m_id

stopifnot(length(evidence(sr2))==2L,any(m_id(e2)==evidence(sr2)))
stopifnot(context(sr2)==context(e2),timestamp(sr2)==timestamp(e2))

ralmond/EABN documentation built on Aug. 30, 2023, 12:52 p.m.