Biology Reference
In-Depth Information
Noncompliance Records
1-Aug
8-Aug
15-Aug 22-Aug 29-Aug -Sep
12-Sep
19-Sep 26-Sep
3-Oct
10-Oct 17-Oct 24-Oct 31-Oct
Microbial Sampling Data
1-Aug
8-Aug
15-Aug 22-Aug 29-Aug -Sep
12-Sep
19-Sep 26-Sep
3-Oct
10-Oct 17-Oct 24-Oct 31-Oct
Figure 9.5
Temporal distribution of regulatory non-compliances and results of testing for Salmonella
observed in one of the U.S. Department of Agriculture-regulated food factories.
the models combined with smart and purposive aggregation of data to pro-
duce informative features, can be the remedy.
An example can be drawn from a recent study into characterizing risk of
a positive outcome of a test for Salmonella performed on a sample of food
taken at a food factory, given the results of regulatory inspections con-
ducted recently at the same factory. Applying regression directly to raw
data does not yield useful outcomes. However, putting the problem into
a temporal context followed by temporal aggregation of evidence, offers
a more useful perspective. Figure 9.5 depicts temporal distributions of a
certain class of failed regulatory inspections (labeled as non-compliance
records) as well as passed and failed (marked with triangles) Salmonella
tests of food samples, obtained for one of a few thousands of the U.S.
Department of Agriculture regulated factories (note: this factory was par-
ticularly prone to violating regulations and was more intensively tested
for Salmonella than the majority of its peers; the typical data from this kind
of a source is sparser).
Let us pick one day in historical data, such as the one marked with arrows
in Figure 9.5. The proposed retrospective data aggregation approach first
checks whether, during a specific period in the future (typically one to a few
weeks), the factory was subjected to any tests for Salmonella in food. If not,
this particular day would be ignored and the algorithm would move on to
the next day. Otherwise, the recent past is checked against regulatory viola-
tions and one of four possible outcomes can occur:
1. True Positive (period with at least one non-compliance followed by a
period with at least one failed microbial test)
2. False Positive (ditto, but all the near-future microbial tests turned
out negative)
3. True Negative (no recent non-compliances followed by negative test
results)
4. False Negative (period with at least one non-compliance followed by
negative microbial tests)
 
Search WWH ::




Custom Search