Mobile Intelligence

11.4 FINDING OTHER INTERESTING PATTERNS

The former section focused on the extension of sequential pattern mining. This section introduces other mining approaches to find interesting patterns from moving object databases.

11.4.1 Spatio-temporal Association Rules

Association rule mining is one of the most popular topics in data mining [1, 25, 56]. An example of an association rule is {notebook} ⇒ {pen}. The rule indicates that if a notebook is purchased, it is likely that a pen is also purchased. So far, various association rule mining methods for large transaction data have been proposed.

Verhein and Chawla [61] extend the notion of association rules to the spatio-temporal context. They call such rules spatio-temporal association rules. The most simple type of rule can be written as

where r_i, r_j are spatial regions and Δ_i, Δ_j are time intervals that satisfy Δ_i < Δ_j. The above rule says that objects appearing in r_i during time interval Δ_i will appear in region r_j during Δ_j. To find interesting rules, they propose the notion of spatial support:

where σ((r_i, Δ_i) ⇒ (r_j, Δ_j)) denotes the conventional support of the rule—the number of objects that satisfy the rule and area(r) is the area of region r. The smaller the area covered, the higher the spatial support of the rule.

In addition, ...