A location-aware GIServices quality prediction model via collaborative filtering

Figures & data

Table 1. Basic symbols and their descriptions in LAGCF.

Symbols	Descriptions
	Target user
	Target GIServices
	QoGIS value of user requesting GIServices
	The time zone which locates in
	The time zone which locates in
	A users set including n users
	A GIServices set including m GIServices
	The set of the common users who invoked both and
	The average value of all the QoGIS value observed by the common users on
	The similarity of the two GIServices and
	A time zone-adjusted PCC similarity
	All the users who invoked the GIServices in the same time zone with the target GIServices
	A subset of users set , in which users both visited the GIServices and
	The users in the same time zone with the target user
	A simplified QoGIS matrix including the top-K similar GIServices and the users group
	The QoGIS experiences of the target user visiting the target GIServices
	A GIServices set including the top-K similar GIServices
	The average of all the QoGIS values which users in visited the similar GIServices
	A user set in which users have never visited the GIServices

Figure 1. Illustration of the design of the LAGCF model.

Figure 2. Description of the historical QoGIS matrix.

Table 2. Description of the QoGIS dataset.

Statistics	Values
Num. of GIServices	3640
Num. of clients	13
Num. of effective QoGIS records	37,593
Num. of involving time zones (clients)	5
Num. of involving time zones (GIServices)	13
Num. of involving countries	24
Minimum response time value	0.01 s
Maximum response time value	174.09 s
Mean of response time value	1.82 s
Standard deviation of response time	3.84 s

Table 3. Fragment of the training QoGIS dataset.

ID	Server URL	Response time (s)	Client IP
…  …	…  …	…  …	…  …
1550	http://Mapdmz.brgm.h/cgi-bin/mapserver	0.47	138.91.114.245
1551	http://gdr.ess.nrcan.gc.ca:80	0.19	138.91.114.245
1552	http://www.opengis.uab.es	0.27	138.91.114.245
1553	http://disc1.sci.gsfc.nasa.gov	1.78	138.91.114.245
1554	http://deli.dnr.state.mn.us	0.51	138.91.114.245
1555	http://54.225.77.204	0.75	138.91.114.245
1556	http://ogc.bgs.ac.uk	0.53	138.91.114.245
1557	http://www.dinoservices.nl	1.41	138.91.114.245
1558	http://geoportal.lneg.pt	4.92	138.91.114.245
1559	http://maps.gns.cri.nz	1.03	138.91.114.245
…  …	…  …	…  …	…  …

Figure 3. The distribution map of the clients and GIServices.

Figure 4. The impact of the matrix density on prediction accuracy.

Figure 5. The impact of the location feature on prediction accuracy.