Page 67 - ITU Journal - ICT Discoveries - Volume 1, No. 2, December 2018 - Second special issue on Data for Good

ITU JOURNAL: ICT Discoveries, Vol. 1(2), December 2018

– Use K-nearest-neighbourhood (Knn) method 4.2.1 Concept implementation in Matlab
to select the most probable pattern candi-
date. In Matlab, the procedural metadata concept is
implemented as a function call. The main code
Then, the data organizer generates, updates, and calling an external function code is considered as a
modifies the tag values, then takes actions or system retrieving and activating procedural
makes decisions based on the procedural metadata metadata from an external repository. By utilizing
as the follows: daily consumption-pattern tags and peak-time tags
– Generate/update the estimated pattern tag of previous days as inputs, the main code performs
and insert the most recent analysis result. pattern estimation based on the pseudo-code
– Organize the tag values and update the data- described in Table 3. Fig. 7 shows a part of the
generated results as an example.
base with changed information.
– Take actions specified in the procedural 4.2.2 Validation on the web
metadata, such as actuate an AMI device to
shut down the electricity, notify the current To show the validity of the proposed concept of
situation to the service providers, or even procedural metadata on the web, the tags used for
call other procedural metadata. the example implemented in Matlab are validated
in web environments. Fig. 8 is the results of pattern
The overall steps of this scenario are described in estimation shown on the web to prove the
Fig. 6. The procedural metadata of this scenario applicability and feasibility of the procedural
analysis is depicted as a pseudo-code since the metadata concept on the web. As mentioned above,
procedural metadata is an executable metadata different types of web data model languages,
coherent with other web technologies, which is not vocabularies, ontologies, etc. can be used for
limited to a single form. procedural metadata as long as they support
micro-formatted tag structures. Therefore, this
4.2 Proof of concept implementation proof of concept implementation can be
generalized for various web environments.
As a proof of concept, the functionalities of the
analyzed scenario are implemented in Matlab, and 5. CONCLUSION
applicability of the procedural metadata concept
on the web is validated. The dataset used in the In this paper, it is argued that establishing common
implementations is real world AMI data metered descriptions of roles of data and devices in smart
for 900 days during 2014–2016 in Gwangju, South systems, specifically in the part of decision making
Korea. It contains 700 low voltage residential users, and performing tasks, is the driver in the
and the electricity consumption load data is development of autonomic and interoperable WoT.
recorded every 15 minutes. To support WoT with common descriptions of logic
and workflows, this paper proposed a concept of
Without processing the original raw data, the procedural metadata. As a type of metadata
implementations use defined procedural metadata containing executable procedures to make
in a form of execution code to perform pattern decisions and perform tasks, procedural metadata
estimation tasks only with the preprocessed provides accumulated knowledge information for
metatags: consumption-pattern tag and peak-time interoperable, autonomic systems. This
tag. The tags are categorized, processed, and information can be shared by heterogeneous
assigned in advance according to several criteria devices and systems which utilize different types
through statistical analysis and machine learning of data formats, languages, vocabularies,
techniques. The consumption-pattern tags are ontologies, etc. as long as they support micro-
acquired through clustering [20]. The peak-time formatted tag structures available on the web.
tags are obtained with a peak searching process. Therefore, heterogeneous IoT environments can be
The entire dataset is organized with proper integrated partially and entirely through the
consumption-pattern and peak-time tags before proposed concept. Hence, procedural metadata
implementation. In the implementation, only the enables automatic and interoperable data, device,
tags are used for electricity consumption pattern and system engagements in WoT. The viability and
estimation. applicability of the proposed concept are shown

© International Telecommunication Union, 2018 45