Table of Contents

Search

  1. Preface
  2. Introduction
  3. Defining a System
  4. Flattening IDTs
  5. Link Tables
  6. Loading a System
  7. Static Clustering
  8. Simple Search
  9. Search Performance
  10. Miscellaneous Issues
  11. Limitations
  12. Error Messages

Reducing Scoring Costs

Reducing Scoring Costs

This section describes about Pre-Scoring.

Pre-Scoring

This feature is used to reduce the cost of scoring two records.
The most expensive part of the matching process is the scoring. Once a set of candidate records is selected each candidate is scored against the search record. In practice, the scoring is often complex in nature and involves several methods scoring several fields on the search and file records.
If it is possible to quickly reject a candidate record by scoring it with a "light weight" (inexpensive) scoring scheme we can avoid the need to call a more complex scheme.
The Search-Definition’s
PRE-SCORE-LOGIC
is used to define an optional "light weight" scoring scheme which is processed before the normal
SCORE-LOGIC
. If it returns a "reject" condition, the more expensive
SCORE-LOGIC
phase is not called.

Scoring Phases

Scoring happens in two distinct phases known as Pre Scoring and Scoring. There are three possible outcomes as a result of the score in each phase. A record may be:
  • rejected, or
  • accepted, or
  • passed to the next scoring phase.
Rejection occurs when the score is less than the
Reject
limit for that phase (limits are built into your SSA-NAME3 population but can be adjusted using the
CONTROLS
parameter in the System definition). A rejected record is not passed to any other phases. It is simply discarded.
A record is accepted when its score is greater than or equal to the
Accept
limit for that phase. It does not participate in any further scoring phases; it is simply added to the search’s result set.
A record that has a score which is greater than or equal to the
Reject
limit and less than the
Accept
limit is deemed to be "undecided" and is passed to the next scoring phase. If no more phases exist the record is returned as part of the search result set.

Adjusting the Accept and Reject limits

Accept and Reject limits are defined by SSA-NAME3 V2 Population Rules. The standard Populations usually define an Accept limit to be less than or equal to 100. If a record reaches a score that is greater than or equal to the Accept limit for that phase, the record is accepted and the record does not take part any further scoring phases as explained above. However, if the reason for using multiple scoring phases is only to reject early and never to accept until the final scoring phase, then the default Accept limit is not useful. In order to never accept records in a scoring phase, one should specify a scoring phase with the Accept limit of 101, which can be never reached. This can be done using the following syntax in the CONTROLS parameter in the Scoring phase definition:
Where AdjA is the Accept level adjustment and AdjR is the Reject level adjustment. To force the Accept limit to become 101 the exact value of +101 must be used as in the following example:
Controls("Purpose=Address MATCH_LEVEL=Typical+101+0")
where the special value +101 is forcing the Accept level to become exactly 101. The Reject adjustment should also be specified. Omitting the Reject limit adjustment means that the Reject level is adjusted by the same amount as the Accept level and the value 101 would cause all records to be rejected, which is clearly not useful. By specifying the value +0 the Reject limit is not changed from the original value specified in the Population Rules.
See the
SSA-NAME3 POPULATIONS and CONTROLS
manual for further details.

0 COMMENTS

We’d like to hear from you!