Mining User Movement Behavior Patterns in a Mobile Service Environment

Mobile service systems offer users useful information ubiquitously via mobile devices. Based on changeable user movement behavior patterns (UMBPs), mobile service systems have the capability of effectively mining a special request from abundant data. In this paper, UMBPs are studied in terms of the problem of mining matching mobile access patterns based on joining the following four kinds of characteristics,
U
,
L
,
T
, and
S
, where
U
is the mobile user,
L
is the movement location,
T
is the dwell time in the timestamp, and
S
is the service request. By introducing standard graph-matching algorithms along with the primitives of a database management system, which comprises grouping, sorting, and joining, these joint operations are defined. Moreover, by mining the associated structure via maximum weight bipartite graph matching, a prediction mechanism, based on the model of UMBPs, is utilized to find strong relationships among
U
,
L
,
T
, and
S
. In addition, a PC-based experimental evaluation under various simulation conditions, using synthetically generated data, is introduced. Finally, performance studies are conducted to show that, in terms of execution efficiency and scalability, the proposed procedures produced excellent performance results.