6. Conclusions
This research examines the issue of developing an effective
service management strategy that Taiwan’s ISP management
is currently facing and proposes a BI process
that could assist management in discovering in-depth
knowledge of customer usage behaviors and network facility
utilization. With IP traffics of each individual user as
data, this process emphasizes significantly on data preprocessing
and subsequent modeling. We present appropriate
technologies along with each phase of the process to aid
management in implementation, which includes a decision
support system to assist management who are otherwise
not capable of integrating various methodologies together.
With the cooperation of a major ISP company in Taiwan,
we demonstrated the detailed implementation of this
process.
The focus of this process is to discover insight knowledge
of users’ network usage behaviors, the characteristics
of their monetary contribution, and facility utilizations; so
that services that are proactive and personalized may be
developed. We applied data warehouse to facilitate the
retrieval of data with different dimensionality, data mining
to discover network behaviors knowledge, and RFM to
characterize users’ monetary contribution. The mining
results of nine customer clusters for the region reveal usage
patterns that were unknown to the management before;
hence this research does present a brand new concept to
the company management in providing personalized services.
These findings identify degree of usage, time of usage, and day of usage of each group and are of important
information to both service department and sales department.
These departments need to re-align resources of both
work force and working hours. The analysis of RFM
model on usage patterns provides further significant
knowledge to management, which could lead them to formulate
proper marketing strategies. For the network
resource management, we apply data mining to find out
network flow distribution among districts of the region,
and map that district flow to the routers that service the
region. Thus, any future investment on router and related
resources will be more effectual and cost effective. For this
particular case, we developed a decision support system
that integrates all methodologies of the BI process
together; so that management could perform analysis when
they need to and on subjects they are interesting in. The
result of system evaluation has indicated very positive
responses from the management staff.