AD ALTA
JOURNAL OF INTERDISCIPLINARY RESEARCH
data, 30% testing data. For the induction of the decision tree, the
library rpart() has been loaded.
For the advanced classifiers more packages have been required.
The 10folds CV requires the e1071 and caret libarires, while the
randomForest library has been loaded to investigate the random
forest classification accuracy. The induced decision (Fig.12) tree
as well as the random forest out of bag (OOB) error plots
(Fig.13) are ilustrated. The OOB estimate relies on observations
which are not considered in the bootstrap sample. The error plot
(Fig. 13) shows that 500 trees were induced for enhancing the
classification robustness. The prediction error is reduced and
stabilized after a specific number of induced trees (<200 trees).
Fig. 12. The decision tree for predicting IVL criticality ranking
of crucial business functions in a public university based on the
full data set (Source: Author)
Fig. 13. The IVL classification out of bag (OOB) error plot
based on the random forest based on the full data set (Source:
Author)
The accuracy of the decision tree, 10-Folds CV and the random
forest algorithms is ilustrated on the following summarized table
(Tab. 4).
Tab.4 Comparison of the predictive accuracy of the three
different classifiers with respect to the recovery priorities for the
entire set of critical business functions in the public university.
IVL Classifier
Accuracy
(based on the confusion matrix results)
Decision tree (CART)
75%
10folds CV
75%
Random forest
91.66%
After considering the above illustrated results, it can be
concluded that the proposed business intelligence tool can be
considered as a rapid and effective software solution for business
continity and safety management. So far, the research has been
focused on the establishment of balanced recovery priorities for
safety critical operations in public organization. However, a
complete safety management policy should also rely on the
robust availability measurement of these operations.
4.4 Computing availability rates for safety-related operations
via RTEmax input for an integrated safety management
framework
Another dimension that must be considered for more effective
safety management policies in public organizations and units is
the estimation of the availability of safety critical operations. An
interesting recent study (Spang, 2017) highlights the importance
of a safety management system in order to protect workers in all
industries form electrical hazards, and defines it as a “formal and
proven system for the safe execution of work activities”.
Moreover seven core safety management principles are indicated
including “balanced priorities” in terms of protecting the
“workers, the public and the environment”. The business
continuity points is proposed a mathematical method for setting
balanced recovery priorities in the occasion of a failure of a
safety critical system and the involved industrial functions.
Additionally, a description of the Integrated Safety Management
system as provided by the U.S. Department of Energy (2008),
indicates, among others, the “operational excellence” as an
important safety management principle. The study relates
operational excellence with high reliability achieved through
“focus on operations, quality decision-making, open
communications, deference to expertise, and systematic
approaches to eliminate or mitigate error-likely situations”.
According to the second principle, the business continuity points
is also aimed to serve as an operational excellence tool for
controlling the reliability and, more precisely the availability of a
safety critical function and the corresponding systems. Based on
the above, the business continuity points can serve as a crucial
part of an integrated safety management under the below
summarized framework (Fig.14)
Fig. 14. Business Continuity Points as Part of an ISM –
Proposed Framework (Source: Author)
Due to the fact that the current version of the developed business
intelligence solution exports business continuity spreadsheet
reports, a visual basic for applications (VBA Excel) software
interface has been utilized for estimating the availability rates for
the selected safety – related business functions based on the
RTEmax vaue that is used. The VBA application has been
developed to estimate availability rates based on the formula (4)
(Rance, 2013) :
100%
ASTDT
Availability
AST
−
=
×
(4)
where, AST = Agreed Service Time, DT=Downtime. DT value
is also mentioned as Mean Time To Repair (MTTR) in other
availability formulas (Garcia et al, 2016)
Assuming that AST=22 hours/day/week (154 hours/week).The
expected downtime will then be, DT=2 hours/day or 14
hours/week.
As a result, the weekly availability for this function is:
A
WEEK
In the case that a maximum unplanned downtime interval of 8
hours/week is permitted the availability rate is then estimated as
= [(154-14)/154] * 100%= 90.9%.
A= [(154-14-8)/154]*100% = 85.71% , which indicates the
maximum tolerable eekly availability rate, so that an
organization will not suffer significantly negative consequences.
Based on the utilized case study, it has been attempted to
estimate yearly availability rates based on the BCPTs proposed
maximum recovery time, with respect to the same safety related
business functions. The following facts have been assumed:
- 363 -