Usage Apriori and clustering algorithms in WEKA tools to mining dataset of traffic accidents

Figures & data

Table 1. Advantages and disadvantages of data mining tools.

Tool	Advantages	Disadvantages
R-programming	Purely statistical, R has better graphics, easier to combine with other statistical calculations	Less specialized for data mining, requires knowledge of array language
WEKA	Easiness of use can be extended in RM. WEKA product gives beginners users a tool to discover hidden information from database and file systems with simple to use options and visual interfaces	Poor documentation, weak classical statistics, poor parameter optimization, weak CSV reader
KNIME	Molecular analysis, Mass spectrometry. Chemistry Development Kit	Limited error measurements, no wrapper methods for descriptor selection, poor parameter optimization
ORANGE	Better debugger, Shortest scripts, poor statistics, suitable for novice experts	Big Installation, limited reporting capabilities
IBM SPSS	SPSS is a good tool for non-programmers. Easy to understand and use, with extensive documentation of use, and is quite strong with quantitative analysis. SPSS is non-code programme	SPSS their graphical results not improving and model intervention is limited

Figure 1. Traffic accident dataset an ARFF file.

Figure 2. Opening ARFF file in WEKA explorer.

Figure 3. Apply the Apriori algorithm.

Figure 4. Apply the EM cluster algorithm in WEKA.

Table 2. Size of a set of large itemsets L (1): 10.

List of attributes	Minimum support	Confidence	Result
1. Year	0.4	0.9	year = 1434  407
2. Time			time = Day  542 time = Night  404
3. Type of accident			type_of_cc = collision_with_other_vehicle  496
4. Location			location = Highway  487
5. Number of vehicles			number_of_v = 2  438 number_of_v = 1  478
6. Driver			driver = both_sides_non_Saudi  380
7. Injured			injured = 0  775
8. Death			Death = 0  911

Table 3. Size of a set of large itemsets L (2): 4.

List of attributes	Minimum support	Confidence	Result
1. Year 2. Time 3. Type of accident 4. Location 5. Number of vehicles 6. Driver 7. Injured 8. Death	0.4	0.9	time = Day injured = 0 Death = 0  447
			type_of_cc = collision_with other_vehicle _number_of_v = 2 Death = 0  406 type_of_cc = collision_with_other_vehicle injured = 0 Death = 0  411
			location = Highway injured = 0 Death = 0  385

Table 4. Size of a set of large itemsets L (2): 13

List of attributes	Minimum support	Confidence	Result
1. Year 2. Time 3. Type of accident 4. Location 5. Number of vehicles 6. Driver 7. Injured 8. Death	0.4	0.9	Year = 1434 Death = 0  387
			Time = Day injured = 0  453 Time = Day Death = 0  525 Time = Night Death = 0  386
			type_of_cc = collision_with_other_vehicle = 2  419 type_of_cc = collision_with_other_vehicle injured = 0  417 type_of_cc = collision_with_other_vehicle Death = 0  480
			Location = Highway injured = 0  392 Location = Highway Death = 0  471
			Number_of_v = 2 Death = 0  425 Number_of_v = 1 injured = 0  383 Number_of_v = 1 Death = 0  459
			Injured = 0 Death = 0  762

Table 5. Best rules using Apriori algorithm.

List of attributes	Best rules found
1. Year	1. time = Day injured = 0  453 ==> Death = 0  447 < conf:(0.99)> lift:(1.02) lev:(0.01) [10] conv:(2.39)
2. Time	2. type_of_acc = collision_with_other vehicle injured = 0  417 ==> Death = 0  411 < conf:(0.99)> lift:(1.02) lev:(0.01) [9] conv:(2.2)
3. Type of accident	3. injured = 0  775 ==> Death = 0  762 < conf:(0.98)> lift:(1.02) lev:(0.02) [15] conv:(2.05)
4. Location	4. location = Highway injured = 0  392 ==> Death = 0  385 < conf:(0.98)> lift:(1.02) lev:(0.01) [7] conv:(1.81)
5. Number of vehicles	5. number_of_v = 2  438 ==> Death = 0  425 < conf:(0.97)> lift:(1.01) lev:(0) [3] conv:(1.16)
6. Driver	6. type_of_acc = collision_with_other_vehicle number_of_v = 2  419 ==> Death = 0  406 < conf:(0.97)> lift:(1.01) lev:(0) [2] conv:(1.11)
7. Injured	7. time = Day  542 ==> Death = 0  525 < conf:(0.97)> lift:(1.01) lev:(0) [3] conv:(1.11)
8. Death	8. type_of_acc = collision_with_other_vehicle  496 ==> Death = 0  480 < conf:(0.97)> lift:(1) lev:(0) [2] conv:(1.08)
	9. location = Highway  487 ==> Death = 0  471 < conf:(0.97)> lift:(1) lev:(0) [2] conv:(1.06)
	10. number_of_v = 1  478 ==> Death = 0  459 < conf:(0.96)> lift:(1) lev:(0-) [-1] conv:(0.88)

Table 6. The distribution of accidents per year.

Year	Total number of accidents	Death	Injured
1432	292	5	58
1434	407	1	81
1435	112	21	22
1436	135	8	10
Total	946	35	171

Table 7. Summarized results obtained by using EM clustering algorithm.

Cluster
Attribute		0	1	2	3	4
Result		(0.29)	(0.1)	(0.44)	(0)	(0.16)
Year	1434	152.0739	10.6103	183.5875	1	64.7282
Time	Day	168.9914	46.8987	245.2619	1	84.848
Type of accident	Collision with other vehicle	2.3265	89.8892	404.5182	1	3.2661
Location	Highway	154.6209	52.2836	186.8669	1	97.2285
Number of vehicles	1	277.908	50.6856	1.7462	1	151.6602
Driver	Both sides non Saudi	142.2255	40.4978	125.3396	1	75.9372
Injured	0	253.7052	66.4617	364.6537	1	94.1795
Death	Death = 0	266	75	421	0	149
	Death = 1	2	3	12	0	13
	Death = 2	0	0	1	0	4