Page 39 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

(a) (b)

Fig 3. Visualized the road segment (a) and accidents (b) based on the dominant land use assigned to each road segment.

2.2. Association rule mining

Association rule is an eminent data mining technique that can reveal interesting relationships between data [10], [18].
Association rules have the form “A →B” (if antecedent then consequent), where A is the antecedent and B is the consequent and
they can have multiple items as antecedent and consequent. ‘Apriori Association rule mining algorithm is used for mining the
frequent patterns in the database. ‘Apriori’ algorithm is used in this research.
In ‘Apriori’ algorithm, the rules were filtered by minsup as minimum support, minconf as minimum confidence, and minL
as minimum lift. Support, Confidence and Lift parameter are generally used to investigate the reliability of the extracted rules.
Support is an indication of how frequently the item set appears in the dataset. To be more precise, the support of → B
rule represents the proportion of transactions in the database that contains the item set A and B ( ∪ ) (Eq. 4) [2].
# ( ∩ )
( → ) = ( ∪ ) = = ( → ) (4)

Where #(A ∩ B) represents the number of crashes when both the condition A and B are verified, and N is the total number
of crashes in the dataset.
Confidence indicates how often the rule is true. Besides, confidence is the proportion of the transactions in the database
which contains ∪ to the number of transaction that contains B ([2],[16]). A higher confidence value for A→ B indicates that
the presence of B is highly visible in the transactions including A. Confidence parameter is calculated based on Eq. 5&6:
( → )
( → ) = (5)
( )

And Lift is calculated as follows:
( → )
( → ) = = ( → ) (6)
( ) × ( )
Fig. 3. Land use attributed to roads (1)- Land use attributed to accident (2)

A lift value greater than 1 indicates positive interdependence between the antecedent and the consequent, while a value
smaller than 1 indicates negative interdependence and finally, a lift value equal to 1 indicates independence ([9], [6]). It is argued
that the high value of support and confidence, and the lift value greater than one prove trustworthy of the extracted rule.

4