Association rule mining for identification of port state control patterns in Malaysian ports

Figures & data

Figure 1. Frequent itemsets generated from the Apriori Algorithm

Figure 2. Mining process for PSC inspection

Figure 3. Malaysian ports

Figure 4. Analysis approach for PSC inspection in Malaysian ports

Figure 5. User-defined threshold based on trial-and-error

Figure 6. Graph for overall analysis using Initial & Follow-up inspections dataset (all data)

Table 1. The rule for overall analysis using Initial & Follow-up inspections dataset (all data)

Rule	LHS		RHS	Support	Confidence	Lift	Count
[1]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.205	0.870	1.706	1659
[2]	{Detention = no, Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.205	0.870	1.706	1659
[3]	{Flag = Singapore}	⇒	{Detention = no}	0.132	1.000	1.012	1065
[4]	{Ship Risk Profile = LRS}	⇒	{Detention = no}	0.236	1.000	1.012	1907
[5]	{Deficiency = zero}	⇒	{Detention = no}	0.510	1.000	1.012	4126
[6]	{Flag = Panama, Deficiency = zero}	⇒	{Detention = no}	0.133	1.000	1.012	1077
[7]	{Deficiency = zero, Ship Risk Profile = HRS}	⇒	{Detention = no}	0.103	1.000	1.012	830
[8]	{Deficiency = zero, Ship Risk Profile = LRS}	⇒	{Detention = no}	0.205	1.000	1.012	1659
[9]	{Deficiency = zero, Ship Risk Profile = SRS}	⇒	{Detention = no}	0.199	1.000	1.012	1611
[10]	{Port = Kuching}	⇒	{Detention = no}	0.124	0.999	1.012	1001

Table 2. Comparison of detention patterns in Malaysian ports

Rule	LHS		RHS	Support	Confidence	Lift	Count
The rule for analysis of detention using Initial & Follow-up inspections dataset (all data)
[1]	{Port = Kuala Terengganu, Deficiency = very high}	⇒	{Detention = yes}	0.001	0.571	48.149	12
[2]	{Flag = Mongolia, Deficiency = very high}	⇒	{Detention = yes}	0.002	0.433	36.513	13
[3]	{Deficiency = very high, Ship Risk Profile = SRS}	⇒	{Detention = yes}	0.003	0.367	30.895	22
[4]	{Deficiency = very high, Ship Risk Profile = HRS}	⇒	{Detention = yes}	0.001	0.345	29.055	10
[5]	{Port = Penang, Deficiency = very high, Ship Risk Profile = SRS}	⇒	{Detention = yes}	0.001	0.312	26.331	10
[6]	{Deficiency = very high}	⇒	{Detention = yes}	0.005	0.211	17.739	40
[7]	{Port = Kuala Terengganu}	⇒	{Detention = yes}	0.002	0.178	15.005	13
[8]	{Port = Penang, Deficiency = very high}	⇒	{Detention = yes}	0.001	0.177	14.949	11
[9]	{Port = Kuantan, Flag = Indonesia}	⇒	{Detention = yes}	0.001	0.177	14.949	11
[10]	{Port = Kuantan, Flag = Indonesia}	⇒	{Detention = yes}	0.002	0.143	12.037	20
The rule for analysis of detention using Initial inspections dataset (detention data)
[1]	{Port = Kuala Terengganu}	⇒	{Deficiency = very high}	0.125	0.923	2.215	12
[2]	{Port = Penang, Ship Risk Profile = SRS}	⇒	{Deficiency = very high}	0.104	0.833	2.000	10
[3]	{Port = Penang, Deficiency = very high}	⇒	{Ship Risk Profile = SRS}	0.104	0.909	1.857	10

Table 3. Analysis of deficiency

Rule	LHS		RHS	Support	Confidence	Lift	Count
The rule for analysis of zero deficiency using Initial & Follow-up inspections dataset (all data)
[1]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.205	0.870	1.706	1659
[2]	{Detention = no, Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.205	0.870	1.706	1659
[3]	{Detention = no, Ship Risk Profile = SRS}	⇒	{Deficiency = zero}	0.199	0.581	1.140	1611
[4]	{Flag = Panama, Detention = no}	⇒	{Deficiency = zero}	0.133	0.580	1.136	1077
[5]	{Flag = Panama}	⇒	{Deficiency = zero}	0.133	0.578	1.134	1077
[6]	{Ship Risk Profile = SRS}	⇒	{Deficiency = zero}	0.199	0.572	1.121	1611
[7]	{Detention = no}	⇒	{Deficiency = zero}	0.510	0.516	1.012	4126
The rule for analysis of zero deficiency using Initial & Follow-up inspections dataset (zero deficiency data)
[1]	{Flag = Indonesia}	⇒	{Ship Risk Profile = HRS}	0.011	0.667	3.314	44
[2]	{Port = Penang, Flag = Singapore}	⇒	{Ship Risk Profile = LRS}	0.024	0.769	1.913	100
[3]	{Port = Kuching, Flag = Vietnam}	⇒	{Ship Risk Profile = SRS}	0.015	0.741	1.898	63
[4]	{Flag = Vietnam}	⇒	{Ship Risk Profile = SRS}	0.035	0.726	1.859	143
[5]	{Port = Lahad Datu, Flag = Marshall Islands}	⇒	{Ship Risk Profile = LRS}	0.011	0.730	1.816	46
[6]	{Port = Bintulu, Flag = Hong Kong, China}	⇒	{Ship Risk Profile = LRS}	0.016	0.698	1.736	67
[7]	{Flag = China}	⇒	{Ship Risk Profile = LRS}	0.023	0.669	1.664	93
[8]	{Port = Penang, Flag = Hong Kong, China}	⇒	{Ship Risk Profile = LRS}	0.010	0.618	1.536	42
[9]	{Flag = Singapore}	⇒	{Ship Risk Profile = LRS}	0.104	0.617	1.533	431
[10]	{Port = Bintulu, Flag = Singapore}	⇒	{Ship Risk Profile = LRS}	0.012	0.612	1.523	49
The rule for analysis of deficiency using Initial & Follow-up inspections dataset (with deficiency data)
[1]	{Flag = Vietnam, Detention = no}	⇒	{Deficiency = low}	0.131	0.886	1.152	521
[2]	{Flag = Vietnam}	⇒	{Deficiency = low}	0.132	0.883	1.148	522
[3]	{Flag = Panama, Detention = no}	⇒	{Deficiency = low}	0.171	0.869	1.130	679
[4]	{Flag = Panama}	⇒	{Deficiency = low}	0.172	0.866	1.126	680
[5]	{Detention = no, Ship Risk Profile = SRS}	⇒	{Deficiency = low}	0.247	0.842	1.095	977
[6]	{Detention = no, Ship Risk Profile = HRS}	⇒	{Deficiency = low}	0.219	0.836	1.087	867
[7]	{Port = Bintulu, Detention = no}	⇒	{Deficiency = low}	0.113	0.823	1.070	447
[8]	{Ship Risk Profile = HRS}	⇒	{Deficiency = low}	0.220	0.820	1.065	873
[9]	{Port = Bintulu}	⇒	{Deficiency = low}	0.113	0.819	1.064	447
[10]	{Ship Risk Profile = SRS}	⇒	{Deficiency = low}	0.248	0.814	1.057	982

Table 4. Patterns discovery at individual ports (Bintulu, Penang, Kuching, Port Kelang and Lahad Datu)

Rule	LHS		RHS	Support	Confidence	Lift	Count
The rule for PSC inspection at Bintulu Port using Initial inspections dataset
[1]	{Flag = Hong Kong, China, Deficiency = zero}	⇒	{Ship Risk Profile = LRS}	0.065	0.698	2.212	67
[2]	{Flag = Vietnam}	⇒	{Deficiency = low}	0.051	0.616	2.102	53
[3]	{Flag = Hong Kong, China}	⇒	{Ship Risk Profile = LRS}	0.073	0.620	1.964	75
[4]	{Flag = Singapore}	⇒	{Ship Risk Profile = LRS}	0.058	0.561	1.777	60
[5]	{Flag = Vietnam}	⇒	{Ship Risk Profile = SRS}	0.051	0.616	1.541	53
[6]	{Flag = Hong Kong, China, Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.065	0.893	1.367	67
[7]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.267	0.846	1.295	275
[8]	{Flag = Panama, Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.065	0.827	1.266	67
[9]	{Flag = Marshall Islands}	⇒	{Deficiency = zero}	0.078	0.816	1.249	80
[10]	{Flag = Hong Kong, China}	⇒	{Deficiency = zero}	0.093	0.793	1.214	96
The rule for PSC inspection at Penang Port using Initial inspections dataset
[1]	{Flag = Singapore, Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.109	0.935	1.424	100
[2]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.350	0.907	1.382	322
[3]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.141	0.812	1.238	130
The rule for PSC inspection at Kuching Port using Initial inspections dataset
[1]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.283	0.953	1.247	243
[2]	{Flag = Singapore}	⇒	{Deficiency = zero}	0.169	0.935	1.225	145
[3]	{Flag = Panama}	⇒	{Deficiency = zero}	0.185	0.832	1.090	159
The rule for PSC inspection at Port Kelang using Initial inspections dataset
[1]	{Flag = Vietnam}	⇒	{Deficiency = low}	0.106	0.882	1.994	75
[2]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.196	0.821	1.669	138
[3]	{Ship Risk Profile = HRS}	⇒	{Deficiency = low}	0.220	0.635	1.435	155
[4]	{Flag = Singapore}	⇒	{Deficiency = zero}	0.112	0.699	1.420	79
[5]	{Flag = Panama}	⇒	{Deficiency = zero}	0.125	0.509	1.033	88
The rule for PSC inspection at Lahad Datu Port using Initial inspections dataset.
[1]	{Flag = Panama, Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.104	0.968	1.204	61
[2]	{Ship Risk Profile = LRS}	⇒	{Deficiency = zero}	0.353	0.954	1.186	207
[3]	{Flag = Marshall Islands}	⇒	{Deficiency = zero}	0.107	0.926	1.152	63
[4]	{Flag = Panama}	⇒	{Deficiency = zero}	0.249	0.849	1.056	146