Computational Intelligence - August 2015 - 64

that actively collects information from
workers and optimizes infrastructure
[10]. The application of predictive cloud
computing technology to large-scale
events such as sports has a rich history
starting with analytics.

100k
90k
80k

3. Sport Analytics

Website Access (HPM)

70k
60k
50k
40k
30k
20k
10k
0k

Fri 00:00

Fri 12:00

Sat 00:00

Sat 12:00

Raleigh, NC, USA
St. Louis, MO, USA
Denver, CO, USA
Figure 1 Hits Per Minute on an anonymized event day at the origin within the 3-sited
private cloud.

Number of Requests

% of Requests

Edge Requests

1.9 Billion

94.4%

Midgress Requests

199,469

Origin Requests

112.8 Million

5.6%

Figure 2 The total number of edge, midgress and origin requests during the same period as
Figure 1.

Other types of elastic resource allocation strategies within the cloud are
reactionary to resource contention. The
Amazon Elastic Compute Cloud (EC2)
combines auto scaling with cloud watch
[2]. Cloud watch publishes usage statistics so that users can change their cloud
configuration through auto scaling
Abstract Programming Interfaces (APIs).
Other reactive commercially available
solutions include Scalr [30] and Rightscale [32]. Autonomic cloud work pro-

vides feedback loops to alter future
cloud resource allocation [29]. Other
works such as Meng et al. and Calheiros
et al. present reaction based systems
[6,23]. A survey on cloud computing by
Galante and Bona classify 28 elastic
computing systems as predictive or reactionary [11]. However, works such Gandhi et al. combined long term predictive
cyclic patterns with reactionary events
[12]. Following autonomic cloud computing, Endo et al. depict an architecture

IEEE ComputatIonal IntEllIgEnCE magazInE | august 2015

Within the market of sports, competitive advantages are driven by data analytics. One of the first papers published
about the application of analytics to the
management of baseball teams was
written by G. R. Lindsay in 1959 [20].
Baseball Prospectus, founded in 1996,
further developed Sabermetrics, or the
analysis of baseball statistics [19]. In
2004, the "Journal of Quantitative
Analysis in Sports" published the first
volume that discussed sports such as ice
hockey, basketball, football, baseball,
cards, and track. Sport analytics began
to become popular with the story of
the Oakland A's in the movie, Moneyball. In 2006, the community began to
mature with the beginning of the
annual MIT Sloan Sports Analytics
Conference. Today, the conference
attracts more than 2,000 attendees.
Tennis and golf embraced analytics
later than baseball and football. One of
the first golfing journals, Science and
Golf, published an analysis of golf swings
and player performance metrics in 1990
[8]. From 1997 to 1999, an academic
statistical analysis performed on all of
tennis Grand Slam tournaments determined the effects of court surface and
player gender on play [25]. In 1999, the
PGA tour added the ShotLink system to
collect data on every golf swing. By
2001, tennis media outlets and fans
applied statistical analysis to tennis players. However, the PCC is a first of a
kind cloud applied within the major
sporting domain, particularly professional golf and Grand Slam tennis.
From a computational intelligence
perspective, several works relate to the
PCC. For example, short-term load
forecasting has been successfully implemented with hybrid filter-wrapper feature selection [38]. Z. Hu et al. use feature selection for their predictive model.
Complimentary, neural networks with

Table of Contents for the Digital Edition of Computational Intelligence - August 2015