Recommending Environmental Big Data Using Semantically Guided Machine Learning - Large Scale and Big Data: Processing and Management - page 463

15.4.5 Unsupervised Knowledge Recommendation .................................... 483
15.4.5.1 Time Series Integration ...................................................... 483
15.4.5.2 PCA-Based Feature Space Representation ........................ 484
15.4.5.3 SOM-Based Big Knowledge Recommendation ................. 487
15.4.5.4 Knowledge as RDF on LOD .............................................. 490
15.5 Conclusion .................................................................................................... 491
References .............................................................................................................. 491

15.1 INTRODUCTION

15.1.1 e nvironmental b ig D ata anD k nowleDge
In information technology, Big Data is a collection of data sets so large and com-
plex that it becomes difficult to process using on-hand database management tools
or traditional data-processing applications [74]. The trend to larger data sets is due
to the additional information derivable from analysis of a single large set of related
data, as compared with separate smaller sets with the same total amount of data
[26,51,68]. Scientists regularly encounter limitations due to large data sets in many
areas, including meteorology, genetics, complex physics simulations, and environ-
mental research [9,35]. Wireless technology-based automated data gathering from
the large environmental sensor networks have increased the quantity of sensor data
available for analysis and sensor informatics. Next-generation environmental moni-
toring, natural resource management, and agricultural decision support systems are
becoming heavily dependent on very large scale multiple sensor network deploy-
ments, massive-scale accumulation, harmonization, web-based Big Data integration
and interpretation of Big Data. With large amount of the data availability, the com-
plexity of data has also increased hence regular maintenance of large-scale sensor
are becoming a difficult challenge. Uncertainty factors in the environmental moni-
toring processes are more evident than before due to current technological trans-
parency achieved by most recent advanced communication technologies [47-49].
The other challenges include capture, storage, search, sharing, analysis, and visual-
ization. Data availability from a particular environmental sensor web is often very
limited and data quality is subsequently very poor. This practical limitation could be
due to difficult geographical location of the sensor node or sensor station, extreme
environmental conditions, communication network failure, and lastly technical fail-
ure of the sensor node. Data uncertainty from a sensor network makes the network
unreliable and inefficient. This inefficiency leads to failure of natural resource man-
agement systems such as agricultural water resource management, weather fore-
cast, crop management including irrigation scheduling and natural resource-based
crop business model systems. The ultimate challenge in environmental forecasting
and decision support systems, is to overcome the data uncertainty and make the
derived output more accurate. It is evident that there is a need to capture and inte-
grate environmental knowledge from various independent sources including sensor
networks, individual sensory system, large-scale environmental simulation models,