Dock Clustering Management System based on Modified K-Mean

Algorithm in Smart Port Services

Ari Wijayanti, Okkie Puspitorini, Nur Adi Siswandari,

Haniah Mahmudah

and

Revfath Risqon Syafaat

Department of Electrical Engineering, Politeknik Elektronika Negeri Surabaya, Indonesia

Keywords: S

mart Port, Lora, Modified K-Means Algorithm, Ships.

Abstract:

In queueing management system, waiting time is a parameter that indicates times duration when the ship

enters the port until it rests on the dock to stevedoring process. T

he problem in waiting time is taking a

long time to

place the ships in the suitable dock for unloaded activity. The condition caused by the ship's

specifications does not match with the specifications of the dock so the ship has to wait for a long. To

overcome this problem, this paper proposed the Modified K-Means Algorithm is used to clustering the

ship into the suitable dock.

The clustering of the ships is based on

Length of All (LOA) or length of the

ship, gross tonnage of ships, and commodity type from ship’s cargo.

Labeling data of dataset according

to the specification of

Port then they are trained using K – Cross-Validation to determine centroid. The

ship's specifications are provided by LoRa (Long Range) device after it is detected first when the ships

enter the port area. The LoRa calculates the distance of each ship using the Euclidean Distance Formula

and grouping the ships based on minimum distance. The experiments have been conducted with 70 ships,

which are clustered by 10 ships on each dock. The result is the accuracy of the Modified K-Means

algorithm in clustering ships were reached 91.4%. This percentage indicates that the ships were

successfully clustering at the correct dock according to the specifications of the ship and the dock

1 INTRODUCTION

As a nation with thousands of islands, ports are

become the most important gate to enter Indonesia

territorial. The main role of

ports is used as an

export and import distribution in the trade sector.

The ports in Indonesia have the weakness lies in the

quality of infrastructure and port management such

as productivity in loading and unloading, severe

congestion conditions, and maintaining old customs

documents (Suprata, 2020). T

he cost of distribution

services by ships will increase as long as the

duration time to park ships over the sea while

waiting for the suitable dock is ready to be entered

(

Nguyen, 2019

). This condition causes the ship

queueing to get longer in the parking area so that

the waiting time increase than before at the port

(Unnati, 2017) (

S.P.Singh

,2013). Such conditions

lead to the emergence of innovation to create a

system to solve this problem. This system serves

to issue a queue for ships to be

anchored so that it is

expected to be able to unravel the queues

of ships

going to the port. Modified K-Means is one of the

simplest and most common clustering methods

that can be used in ship queue management. It can

group large amounts of data with relatively fast

and efficient computation time. However, The

results of the cluster formed by the K-means

clustering method are very dependent on the value

of the initial point cluster initiation.

Therefore,

the

Modified K- Means

algorithm is proposed to

counter the problems. This algorithm is following

predetermined parameters as a previous study in

research (Emre,2011) (Oyelade,2010).

By implementation of Modified K- Means

algorithm in

queueing management system, the

number of waiting times due to misplaced ships

will be overcome and more coordinated to optimize

port performance. In previous studies conducted,

the vessel detection process was carried out with an

ultrasonic sensor, where this sensor has a limited

range to detect vessels (Swapna Ch, 2017) (A

Kamalov, 2019) The weakness is the process of

detecting objects tends longer and there is no port

692

Wijayanti, A., Puspitorini, O., Siswandari, N., Mahmudah, H. and Syafaat, R.

Dock Clustering Management System based on Modiﬁed K-Mean Algorithm in Smart Port Services.

DOI: 10.5220/0010951200003260

In Proceedings of the 4th International Conference on Applied Science and Technology on Engineering Science (iCAST-ES 2021), pages 692-697

ISBN: 978-989-758-615-6; ISSN: 2975-8246

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

clustering process first, cause system performance

to be less than optimal in determining the port for

ships (Unnati, 2017) (

S.P.Singh

,2013). In this

research,

LoRa (Long Range) sensors are used to

detect the arrival of ships into the port at the first with

their specification, assumed by a normal condition

in the sea environment. A

t the port gate, the system

clustering them and classify the suitable dock by

implementing the modified K- Means Algorithm.

It affects the efficiency of placement time in the

fastest matching process than before so that the

determination of the destination of the ship's dock

will be faster and reduce the waiting time list. All

of the data in this research are based on Tanjung

Perak port container terminal.

This paper is organized as follows: section 1

introduction, section 2 is the theory, Section 3 is

the system design and section 4 is the conclusion.

MODIFIED K

MEANS

ALGORITHM

Modified K - Means Algorithm is one of the most

popular algorithms used for clustering data because

it can be easily implemented and is the most

efficient in terms of execution time. With this

algorithm, data of a similar type is tried to be

grouped from large data sets carried out by repeated

calculations. As a result, the computational

complexity of this

algorithm is very

high

(Oyelade,2010).

Several studies have been

conducted to minimize this, K-Means algorithm is a

sensitive

algorithm because the process must

determine the correct centroid value as a reference

cluster. The error that often occurs in this algorithm

is when determining the first

centroid, if the first

centroid is not appropriate then the cluster results

obtained are also not optimal or even an empty

cluster.

To calculate the distance between data and

centroids, the K- Means algorithm uses the

Euclidean distance formula

(Sharfuddin,2015)

K-Means Algorithm is used to clustering data

by calculating the closest distance between the data

centers determined by each data, the following step

K - Means algorithm works in clustering data

(Vaishali,2011):

Determine the number of K clusters.

Determine the centroid randomly

Calculate the distance of each data using the

Euclidean distance formula

Clustering each data by clustering

according to the minimum result of the

Euclidean distance from each centroid

Compute new centroid from new clusters

Iterate until a convergence condition occurs

Convergence condition is a condition where

the members of each cluster do not change after an

iteration. The following Euclidean distance formula

is used to calculate the distance of each data by

centroid as in equation(1)

(M Emre Celebi,2011)

Euclidean Distance =

√

∑

(

𝑃𝑘

−

𝑄𝑘

)

(1)

Where, n = Total Data,

Pk = Value of Centroid

Qk= Value of Each Data

2.1

K – Cross-Validation

K-Fold Cross Validation is primarily used in

applied machine learning to estimate the skill of a

machine learning

model on unseen data. That is, to

use a limited sample

to estimate how the model is

expected to perform in general when used to make

predictions on data not used during the training of

the model. An example of applying k-fold cross-

validation is shown in

Figure 1: Example of cross k-fold validation.

In this research, the K-Means Algorithm is used

for the clustering process in the dataset. Some

modifications are made to the K- Means Algorithm,

in addition to determining the random centroid

value in the original K-Means Algorithm, which

can cause suboptimal clustering results, so that the

centroid generation section also requires

modification, namely the labeling of the dataset. In

this research, the dataset comes from Pelindo based

on the specifications of the port and ships. First,

it has trained the dataset using K-Cross-Validation

Method to find out the

accuracy of the centroid value

which is generated later after conducting the data

training process. Later the centroid generation

process is done with the python program.

The following step Modified K - Means

algorithm works in clustering data.

Input Dataset from Pelindo

Labeling Data of Dataset according to

Specification of

Port

Dock Clustering Management System based on Modiﬁed K-Mean Algorithm in Smart Port Services

693

Training Data Using K – Cross-Validation

Determine Centroid

Calculate the distance of each data using

Euclidean Dist ance Fo rmula

Grouping based on the minimum distance

3 SYSTEM DESIGN

The overall design of the system starts when the

ship entering the port, the system will determine the

ship's dock to carry out the loading and unloading

process. For the clustering process, the system will

use Modified K-Means Algorithm with parameters

that have been obtained from the database system.

The parameters used by the system are LOA and

type of commodity. Clustering will prevent the ship

from parking its ships that are not following the

specifications of the ship and dock. After that, the

scout ship will pick up the ship to carry out the

loading and

unloading process at the specified dock.

For the entire scheme

and flowchart system shown

in Figure 1 and Figure 2.

Figure 2: System Scheme.

From the flowchart in

Figure 3, The first thing the

system does is collecting data collection, where this

data collection is a reference

for the algorithm that

the system is used.

The ship will enter the port’s gate

after it

receives the dataset from LoRa sensors. The

port will be clusterized using the k-means algorithm

with the ship length

parameters and the types of

commodities that have been adjusted to the

previous data. At the end of the process is the ship

gets information about a suitable dock for

unloading the cargo.

Figure 3: Flowchart System.

3.1 Data Collection

The data that is used as a

reference for the Modified

K-Means Algorithm in clustering

ships is as

follows:

•

Specification of Ship

•

Specification of Dock

•

Dock Operational Time

•

Dock Facilities

•

Dock Service

•

Ship loading and Unloading Mechanism

These parameters are needed by the system as a

reference

in clustering. It

is needed by the system as

a reference

in clustering. Dock facilities and

services are important data to match the utility of

the ship.

Specification of the ship is primary data to

seek a correct dock for parking time as long as the

unloading cargos process. During the process, the

management pays attention to dock operating time

to inform the gate. This part is an important step of

the ship loading and unloading mechanism.

Some specifications of the ship that K-Means

Algorithm needed to classify ship according to the

dock as follows: Length of All (LOA) of Ship means

the size of the ship and the commodity types of ship’s

cargo. Based on the data of the Tanjung Perak port

management, the dock types are divided into some

utility such as for passenger, dry bulk, liquid bulk,

and general cargo named. The Specification of dock

services Port as shown in Table 1.

iCAST-ES 2021 - International Conference on Applied Science and Technology on Engineering Science

694

Table 1: Specification of dock services at Tanjung Perak

port.

Dock

Commodity Services

Jamrud Utara

Passenger

Dry Bulk

General Cargo

Jamrud Barat

General Cargo

Dry Bulk

Jamrud Selatan

General Cargo

Kalimas

General Cargo

Mirah

Liquid Bulk

General Cargo

Container

Berlian Timur

Container

Nilam Timur

Liquid Bulk

Dry Bulk

General Cargo

Container

Table 2: Dock Dimension in Tanjung Perak Port.

Dock

Length (m)

Width (m)

Jamrud Utara

1200

Jamrud Barat

217

Jamrud Selatan

800

Kalimas

2270

Mirah

640

Berlian Timur

780

Nilam Timur

920

Figure 4 shows the flowchart of the Modified

K-means algorithm for clustering ships. The

clustering

process of ships needs the dock dimension.

Each ship will be matched first in the dimension of

the dock before continuing to commodity types of

cargos. Loading and unloading cargos mechanism are

done inside the suitable dock according to a specified

time. Table 2 shows dock dimension in length and

width.

Figure 4: Flowchart Modified K-Means Algorithm to

Classify Ship.

3.2

Training Data Process for Ship

Clustering

In the K-Means algorithm, there is a step that

determines the centroid value, for this reason, the

data training stage is performed first to see the

accuracy of the centroid value generated later. The

training data is carried out using the k- cross-

validation method to determine the accuracy of the

centroid value that will be generated later.

After training as many as 300 data with the K-

Means Algorithm to classify port according to ship

specifications and use 21 tests with a comparison

of the number of a different number of k-folds. In

figure 5 shows the process of taking data sets of

300 data

from each dock with their

commodity has

been obtained from Pelindo III Tanjung Perak port

container terminal Surabaya.

Figure 5: Result of the training data process.

In the process of training the K-Means

algorithm in clustering port in Table 3, after testing

21 times where each test changes the value of

Dock Clustering Management System based on Modiﬁed K-Mean Algorithm in Smart Port Services

695

k_folds where this value is very influential in

seeing the accuracy of the training process and it is

found that the average accuracy of 84.658 percent

with a data set of 300 data set. Later in generating

centroid values that take from the Port data and

which will be made the partisan k-means algorithm

in determining ship clustering will get a k fold

accuracy average of around 84.658%. The

percentage means the accuracy is still high and

good. According to this process, K-Means

Algorithm can be implemented in this system

especially in the ship clustering process.

Table 3: Result of training data for ship clustering.

Dataset

K_fold

Accuracy (%)

300

74.667

300

87.667

300

85.926

300

88.333

300

88.333

300

85.926

300

105

80.952

300

120

300

135

85.926

300

150

300

165

81.212

300

180

80.556

300

195

300

210

81.429

300

225

84.444

300

240

85.417

300

255

85.490

300

270

85.556

300

285

86.667

300

88.333

Average

84.658

3.3

Centroid Generation Process

For the generation of centroid values for each dock

and commodity using the python program with the

dataset from Pelindo, the results are as shown in

Table 4. This centroid value will be the reference for

the k means algorithm in clustering ships according

to the specification of ships and the specification of

each dock. The graph for the value of centroid for

each dock and commodity is shown in figure 6.

Figure 6: Value of Centroid for Each Dock – Commodity.

Table 4: The results of the centroid generation process.

Doc

Commodity Services

Centroid

Jamrud Utara

Passenger

122.9

Dry Bulk

175

General Cargo

110

Jamrud Barat

General Cargo

134.5

Dry Bulk

150.9

Jamrud Selatan

General Cargo

91.9

Kalimas

General Cargo

52.7

Mirah

Liquid Bulk

69.5

General Cargo

75.5

Container

106.6

Berlian Timur

Container

131.1

Nilam Timur

Liquid Bulk

51.3

Dry Bulk

115.2

General Cargo

62.6

Container

125.9

4 RESULT

After conducting experiments with 70 ships,

which 10 ships every dock at Jamrud Utara,

Jamrud Selatan, Jamrud Barat, Kalimas, Mirah,

Berlian Timur and Nilam timur.

Table 5: Result for ships clustering for each dock.

Dock ships Success

Fail

Accuracy

(%)

Jamrud utara

100

Jamrud barat

Jamrud Selatan

Kalimas

100

Mirah

Berlian Timur

Nilam Timur

Average

91.4

This success statement indicates that the ship was

successfully clusterized at the correct dock according

iCAST-ES 2021 - International Conference on Applied Science and Technology on Engineering Science

696

to the specifications of the ship and also the dock,

while the failed statement indicates that the ship was

not successfully

classified correctly.

Based on Table 5, it can be seen that in the

clustering process of 70 ships with 10 ships per

dock get different results, at the Jamurd Utara and

in Kalimas all 10 ships have been correctly

classified so that they get an accuracy of 100%,

then at the Jamrud Barat, Jamrud Selatan, Mirah,

and Berlian Timur every 9 ships were successfully

classified and only one ship failed to be classified

correctly to get an accuracy of 90%, then at the

Nilam Timur, 8 ships are classified correctly so

that only get an accuracy value of 80%, from all

dock the average gets an accuracy of 91.4% Graph

results of Ships Clustering shown as figure 7

Figure 7: Result accuracy of Ships Clustering.

In the clustering process using modified K-

Means algorithm, there are some errors in

clustering ships, this is because the centroid value

generated between the same commodity with

different dock has almost the same value or there is

no significant difference in value, this causes errors

in clustering ships, especially dock with the

specification that has the same type of commodity.

5 CONCLUSIONS

This paper, concludes that the Modified K-Means

Algorithm clusterized the ships' accuracy to 91.4%

overcome placement errors that exceed the value

of residence

time by using LOA parameters and

commodity types from the ship specifications. By

clustering the ship by the specified dock, a high

waiting time value caused by incorrect placement

of the ship can be reduced appropriately, so that it

can optimize the performance of the port. The

future work it is desirable to have higher accuracy

by applied and combine with other algorithms.

ACKNOWLEDGEMENTS

Thanks are due to Maritim teams Revfath Risqon

Syafaat, Fahmi Nurdin Handy Novian, Dimas

Khrisna Ramadhani. and

PELINDO III

Surabaya

for your cooperations

REFERENCES

Unnati R. Raval and Chaita Jani (2016). Implementing &

Improvisation of K – Means Clustering Algorithm.

International Journal Computer Science and Mobile

Computing, vol. 5, issue. 5, pg. 191-203

Dr. S. P. Singh and Ms. Asmita Yadav (2013 ). Study of K

– Means and Enhanced K – Means Clustering,

International Journal of Advanced Research in

Computer Science. vol 4, No 10

M Emre Celebi(2011). Improving the performance of K-

Means for Color Quantization. Image and Vision

Computing. Vol 29(4).260- 271.

Oyelade, O.J, and Oladipupo O.O and Obagbuwa I.C

(2010). K – Means clustering algorithm for students

academic performance. International Journal of

Computer Science and Information Security. vol 7, No 1.

Sharfuddin Mahmood (2015). A Proposed Modification of

K-Means Algorithm. International Journal of Modern

Education and Computer Science, Vol 6, pp 37-42

Vaishali R.Patel (2011). Modified K-Means Clustering

Algorithm. Computational Intelligence and

information technology (CIIT). pp 307-312.

Sk Ahammad Fahad,” A modified K-Means Algorithm for

Big Data Clustering (2016). International Journal of

Computer Science Engineering and Technology, Vol 6,

Issue 4,129-132

Swapna Ch, Mukesh KumarNaswal Kishor Y and Niraj

(2017). ”tracking and emergency detection of inland

vessel using GPS-GSM System, international

Conference on Recent Trends in Electronics,

Information and communication technology (RTEICT)

Ahmad Kamalov and suhyun Park (2019). An IoT based

Ship Berthing Method using a Set of Ultrasonic

Sensors. MDPI jurnal sensors 2019. 19,5181;doi:

10.3390/s19235181

Minh-Duc Nguyen, Sung-june Kim, 2019, An Estimation

of the Average Waiting Cost of Vessels Calling

Container Terminals in Northern Vietnam, Journal of

the Korean Society of Marine Environment & Safety,

Vol. 25, No. 1, pp. 027-033, February 28, 2019, ISSN

1229-3431(Print) / ISSN 2287-3341(Online)

F. Suprata (2020), analysing the cause of idle time in

loading and unloading operation at indonesian

international port container terminal: port of tanjung

priok case study. Iop Conference Series materials

science and engineering 847 012090, doi: 10.1088/

1757-899x/847/1/012090

Dock Clustering Management System based on Modiﬁed K-Mean Algorithm in Smart Port Services

697