AD ALTA
JOURNAL OF INTERDISCIPLINARY RESEARCH
each business function is highly critical, AST=24
hours/day (8760 hours/year),
the DT value has been replaced by the RTE
MAX
the DT value has been replaced by the maximum tolerable
downtime proposed by the university BCM team, in order
to compute the proposed by the university weekly
availability rates and compare them with our computed
results,
proposed
by the BCPTs tool to compute the BCPTs weekly
availability rates,
one outage incident per year is considered, the duration of
which is RTE
MAX
For the above mentioned computations a simple VBA excel tool
has been developed and utilized (Fig. 15).
.
Fig. 15. The VBA Excel form for estimating availability rates for
individual business functions based on recovery time input
Tab.5 Predicted availability rates for safety related university
business functions.
Business
Function
Proposed
Maximum
Recovery
Time by
the BCPTs
BI tool
(hours)
Proposed
Maximum
Downtime
period
from the
university
BCM
team
(hours
Predicted
Yearly
Availability
Rate based
on RTE
MAX
Proposed
Yearly
Availability
Rate by the
University
BCM
experts
(BCPTs
approach)
Emergency
Communication
24
24
99.72%
99.72%
Public
Information
24
48
99.72%
99.45%
Risk
Management
168
24
98.07%
99.72%
Police and
Security
72
48
99.17%
99.45%
Mail Services
24
24
99.72%
99.72%
Core
Technology
Infrastructure
72
72
99.17%
99.17%
Emergency
Services
24
24
99.72%
99.72%
The conducted results indicate satisfactory performance levels
for both the BCPTs approach and the proposed BI tool in terms
of availability estimation for safety critical operations. It can be
noticed that only for two business functions, taht is, the risk
management and the police and security the computed
availability slightly deviates from the proposed by university
BCM experts rates. However, 5 out of 7 estimations were highly
accurate, while in the case of the public information function the
proposed by our approach availability rate was higher than the
one proposed by the BCM experts.
5. Conclusions and future research directions
Business intelligence solutions are important for ensuring
resilinece in public organizations. The big data manipulation is a
modern challenge of paramount importance for business
continuity and safety management. Public universities include
several safety-related business functions for which rapid
restoration after unexpected interruptions and high availability
rates are crucial for their smooth operation. The current work has
been focused on the development of a modern business
continuity and safety management tool based on the business
intelligence data warehouse concepts. The dimensions, facts and
the defined information granularity has been designed based on
the business continuity points (BCPTs) method, that is utilized
for the proactive recovery priority level definition as well as the
proactive computation of the resumption timeframe for
individual business functions. The method is based on the
computation of recovery complexity and effort estimation
parameters which have been inspired by the Use Case Points
approach. In the present article, the BCPTs approach has been
validated via a real data set from a public university. The
proposed tool is supported by a web interface which facilitates
the BCPTs computations and enables the estimation of the
criticality ranking and the recovery time effort estimation. The
exported spreadsheet data can be used for OLAP operations and
data mining activities. From the utilized data set which is
composed of 42 university business functions we selected 7
safety related functions to measure the accuracy of the BCPTs
classifier and the computations conducted by the proposed BI
solution. The entire dataset has been also investigated with
machine learning classification techniques, namely the decision
trees, the 10Folds cross validation and the random forests with
respect to the IVL predictive accuracy. The estimated accuracy
in predicting the critility level (impact value level - IVL) has for
the safety-related operations has been 85.71% which is highly
promissing. For the full dataset the decision tree classification
and the 10folds cross validation techniques were 75% accurate
based on the confusion matrix results stemming from the testing
data (30%) over the full dataset. The random forest technique
was 91.66% accurate. Finally, the safety-related functions, have
been used for investigating the accuracy of the current tool in
estimating their availability rates. The availability rates
computed via the suggestd BI tool, have been highly accurate
when compared to the proposed availability rates by the
university BCM experts. A developed VBA Excel simple
interface has been used to support availability computations
which stem from the exported spreadsheet data. Only 2 out 7
business functions slightly deviated from the proposed by
experts rates. In general the results conducted throughout the
present resarch are highly encouraging for the proactive business
continuity and safety management in public organizations and
especially for universitities which has been the target domain of
the present article.
Nevertheless, crucial future research activities include the further
validation of the BCPTs BI solution by extracting data from
more universities, the incorporation of more data features in
order to infer more advanced machine learning classification
techniques and, also, the enrichemnt of the current web-based BI
interface by including more functionalities. One of them is the
connection of the application with sophisticated machine
learning software packages such as the R package. However
larger data volumes from other universities should be gathered.
This task is demanding due to the fact that BCM data is sensitive
and confidential in most cases. However, data from 5 more
universities which is currently processed and analyzed by the
research team have been so far collected and will be used for
future investigation. Moreover, the incorporation of the current
standalone VBA tool in the BI solution for computing the
availability rates is planned. Finally, the current interface
requires further testing for estimating accurate resumption
timeframes based on several recovery scenarios for every
individual function. In this way, machine learning regression
tasks can be performed. Similar BCM BI solutions can be also
proposed for other public organizations.
- 364 -