The Identification of Typologies and Levels Utilization of Non-Commercial Ports in Indonesia Using the Machine Learning Method

A port is essentially an area designated as an arrival point for the loading and unloading marine freight [9]. Seaports are the most practical option for loading and unloading watercraft, ships, ferries, and other terminal activities. Ports are always near the water to allow interaction between land and water transportation activities and accommodate vessels and other water vehicles [10]. Ports are business and production facilities that contribute to a country's socioeconomic development, influenced by adjustment in society and economy characterized by required expansion in the nations through nation-state involvement to improvements in logistics, economics, organizations, and the setting of commerce liberalization technology. The power and size of modern ships obscure their early duty. The port enables freights and passengers to move from one kind of conveyance to another, from a vessel to land transportation or to the other way [11]. Seaports as vital logistical hubs for global trade. As a result, because all maritime enterprises must pass via their facilities, where the method of transportation is changed from water to land, seaports are crucial to the logistics chain [12]. The seaport serves as a gateway for connecting areas and nations as one of the maritime transportation infrastructures. Indonesia's port infrastructure is the most critical in defining the country's economic path. Indonesia depends heavily on port reliability to enable upward mobility and inter-island trade because Indonesia is an archipelagic country [13]. Ports are territories connected by a waterway that connects to the ocean's surface, sea, or stream as entities. They have the necessary infrastructure and technological capabilities to handle the cargo type. Its primary objective is to give various levels of protection to ships, allowing goods transporting from one mode of transportation to another [14]. A Port is also defined as maritime infrastructure that provides access to seagoing ships [15].

The port is one of the most essential trade networks in the commerce process for inter-island and worldwide trade. Harbor is a place to carry out activities of moving goods from one place to another that transporting via sea transportation routes, where the process originates from origin to destination. The function of the harbour can be mentioned as a meeting place (interface), gateway (gateway), industrial entity (industry entity), and a meeting place for various modes of transportation. Four port functions consist of Gateways; The port is called the point of entry because it is an official area for trade activity. Customs and quarantine procedures must be followed at the points of entry and exit of goods, so a process has been arranged at the port. Link (chain), The presence of a seaport essentially promotes cargo transfer between forms of land transportation (inland transport) and modes of marine transportation (maritime transport), distributing products as fast and efficiently as feasible in and out of customs areas. Gateway is a country or location because a ship can enter a country/region through the port of the country or region concerned. The interface is meant by interface here: in the flow, the distribution of goods inevitably has to pass repeatedly across the port area, once at the points for loading and once at the unloading ports.

If the port is held well, this industry entity will expand and develop additional commercial categories, transforming the port region into an industrial port zone, which will grow shipping companies engaged in the field, agency, warehousing, trucking, etc. The port role has progressed from the conventional procedure of assisting as an additional detailed link comprehensive supplies network as part of global distribution systems, with transportation and loading activities [16]. Roa Perera et al. [14] divides port criteria into several types based on the basin's depth and the cargo amount. Their research defines the seven port types: Port Terminal, Pier, Deepwater Seaport, Harbor, Off-shore, Canal, and River Port, according to the amount of cargo. Fonseca et al. [17] define ports as minor, load points, and regional and international pivot ports. These ports are categorized based on their position, hinterland, and capacity.

Various countries implemented different categorization schemes in port segregation and have been combined to generate the widely known port classification schemes. The various categorization standards utilized by numerous nations in port segmentation were merged to create the widely known port classification frameworks. For example, Ding et al. [18] classified Ireland ports as small, medium, and large ports performing yearly freight volumes of over one million tonnes, medium-sized harbours managing every year shipment amounts of between one hundred and fifty thousand and one million tonnes, and tiny ports handling annual cargo volumes of less than one hundred and fifty thousand tonnes [18]. In contrast, administrative significance was used to classify India's ports as significant, intermediate, and small. Temporal cargo volume changes were used instead, providing a consistent basis for several applications. The administration obtains port traffic data and plans, designs, and operates port infrastructure [19]. Based on the Department of Transport [20] the major ports in the United Kingdom handle at least one million tonnes of cargo annually, while the minor ports handle less than one million tonnes of cargo annually.  The EU statistics taxonomy categorizes port sizes based on the port area's dimensions, primary activities, and port company operations. Small ports manage less than or equal to 10 million tons of cargo handled yearly, while medium-sized ports manage more than ten and up to 50 million tons of freight annually [21]. Ports must carefully analyze the governance model best suited to their unique needs and circumstances in order to maximize performance, and they must periodically review and evaluate their governance structures to make sure they successfully support the port's operations [22]. The World Bank [23] states that port management responsibilities can be modeled in four different ways. The management model is different from labor, asset ownership, and port service delivery.

Indonesian National Port System is a system that includes responsibilities, operations, categories, and structure, a national port strategy plan, and port sites, as well as intra- and intermodal connectivity and collaboration with various sectors. It describes port development based on economic area, geography, geographic region comparative advantage, and natural conditions. National Port Master Plan grouping ports into Main Ports, Collecting Ports, and Feeder Ports. Main Port is a port whose primary purpose is to handle activities such as national and global maritime transportation, the transhipment of national and worldwide marine freight in large numbers, as the starting point of travellers and commodities, and also for ferry transport with regional area service coverage. Collector Port is a port that's primary objective is to provide local sea shipping operations, such as transhipment for medium local movement and as the origin of persons or commodities, as well as ferry transport with a regional service area. A feeder port is primarily responsible for handling national maritime transport and inadequate local maritime transshipments, is a feeder for the main port and collection port, and serves as the entry point of people and goods alongside other transit intersections with service coverage within the province. Indonesian ports are divided into commercial and non-commercial ports based on how they are managed. Commercial ports are operated by the private sector, namely port business entities, while non-commercial ports are governed by the government, in this case, the Directorate General of Sea Transportation.

It is necessary to raise the non-commercial port utilities still-poor performance. Port performance is evaluated using several types of criteria. Vaggelas [24] lays established standards for evaluating port performance, such as expenses, operation time, reliability of infrastructure, connectivity, and access, and ability to operate when needed. The primary elements that serve as criteria for service users in choosing a port are satisfaction, flexibility, price, and service time. The results of port performance measurements can be used as a guide in increasing port competitiveness [25]. Mickiene and Valioniene [26] assesses the effectiveness of government-owned ports and the national economy's effects on port operations. Today, ports serve as the hubs of international trade, provide value-added services to various stakeholders, and serve as meeting places for carriers and shippers. So, the idea of port performance has evolved significantly over time. The methods for evaluating port performance were divided into three categories by Bichou [27] performance measurement, economic implications, efficient frontier approach, and technology capability. One of these ways highlighted the port's service performance. Many scholars believe that the infrastructure's length, design, and upkeep are essential in determining the port's infrastructure and superstructure's quality.

Earlier studies concentrated on port revenue and effectiveness operational issues since they substantially impact a port's competitiveness [28]. The average performance at terminals/berths was utilized by Cabral and de Sousa Ramos [28] to evaluate the performance of Brazilian seaports. Rozar et al. [29] make their selection based on competitiveness criteria like improving the efficiency of material handling facilities, improving flexibility (loading), lowering the cost of loading and unloading, improving labour productivity, improving the location of stockpiles, increasing trucking productivity, reducing the lead time for loading, enhancing service quality, and lowering the processing time for loading and unloading.

According to the time ship services are provided while in port, the port's efficiency can be utilized to gauge the service offered to users (ships and goods). The high port performance demonstrates the port's capacity to deliver good services. There is a close relationship between port efficiency and other port functioning parameters. Thus, the performance of the port affects its efficacy [30]. In realizing a sustainable port, port performance assessment is not limited to its productivity but must consider social and environmental factors (contribution to the surrounding community). Port performance indicators are significant for port authorities. They can be utilized to enhance port services and act as a priority foundation for future port expansion planning [7].

Data mining is the procedure of obtaining information from enormous quantities of data. A method called clustering divides data into groupings so that data in a group has a uniform type. With this technique, data can be grouped or class clustered, where objects with something in the joint can be compared to other things. Classification algorithms in data mining are capable of managing massive amounts of data. It can form hypotheses about categorized class names, arrange information based on training sets and class labels, and identify newly available information. The K-means algorithm is an algorithm that can classify what types and groups of data it is based on the centroid point closest to the data information. This K-means algorithm aims to group data by striving for similarity or similarity of information data in clusters. Rozar et al. [29] apply cluster analysis to evaluate the level of competition in Malaysian ports and divide 18 Malaysian ports into two distinct classes based on various performance metrics. Azzam et al. [31] determine a new port category system based on port qualities and features from a strategic standpoint. The K-means clustering technique was used to carry out the port classification process at many stages. Twenty-five characteristics with 62 elements were found and used to evaluate and categorize ports worldwide. Classification seeks to anticipate the target class as accurately as possible. During training, the classification technique discovers the relationship between the supplied attribute and the outcome characteristic to develop a model. Data mining scenarios generate a lot of information. The decision tree method works great if the data collection is well-categorized and has few nodes. Each path in a decision tree, beginning at the root, is characterized by a data-separating process until a Boolean value is created at the next branch point. It is a stacked arrangement of information connections via linkages and nodes. When using relations for the organization, nodes reflect purposes. A decision tree is a technique for supervised machine learning for problem-solving in classification and regression. Iteratively, decision trees are constructed by applying split criteria at each node to divide the training records into subsets with an outcome variable. Ghiasi et al. [32] developed a decision tree to classify models for medical diagnosis. Tangirala [33] applied the Decision Tree Classifier Algorithm to assess the impact of the GINI Index and Information gained on classification. Kang er al. [34] use a red tide surveillance approach utilizing decision tree classification to more correctly identify the red wave prevalence area.

Orange is a data mining software that is organized hierarchically. Lower-level operations, such as chance examination, data screening, and attribute assessment, are incorporated into more complicated algorithms, such as classification tree learning techniques. Data Analysis and preparation includes data cleaning and selection, inference, feature manipulation (discretization, continuation, normalization, scaling, and scoring), choosing attributes, and categorization using different supervised machine learning algorithms (tree, forest, instance-based, and Bayesian approaches), and regression with lasso regression Orange's script library serves as the foundation for the company's visual programming platform, which includes interface components engaging data visualization using graphical users [35]. Machine learning, used in data science, is a computer programming process that learns from previous experiences. Machine learning aims to create algorithms for summation that learn directly or indirectly from data. Machine learning algorithms may predict, classify, and make decisions, among other things. Machine learning is an artificial intelligence subject study that has grown in popularity important aspect of data science. Orange's research will be used as part of the data mining process, with the primary objective being to obtain the most efficient and effective design determined by the algorithm's examination results. Machine learning starts with constant training data as input. Clustering ports using machine learning techniques can significantly enhance productivity and efficiency in port operations. By leveraging advanced algorithms, such as k-means clustering or hierarchical clustering, on vast datasets encompassing port characteristics, operational metrics, and environmental factors, stakeholders can identify patterns and relationships among ports that were previously obscured. This data-driven approach allows for the creation of clusters based on similarities in infrastructure, cargo handling capabilities, geographical location, and other relevant attributes. Once ports are grouped into clusters, decision-makers can gain valuable insights into best practices, performance benchmarks, and areas for improvement within each cluster. By analyzing cluster-specific trends and patterns, port authorities and operators can tailor their strategies and investments to address specific needs and challenges, thereby optimizing resource allocation and enhancing operational efficiency [36]. Port clustering aims to minimize competition by promoting collaboration, specialization, and synergy among neighboring ports. Instead of viewing nearby ports solely as competitors, clustering encourages them to cooperate and complement each other's strengths, thereby maximizing collective efficiency and effectiveness. By clustering ports based on shared attributes, such as geographic proximity, market specialization, or infrastructure connectivity, stakeholders can identify opportunities for collaboration and resource-sharing that benefit all ports within the cluster. Port clustering minimizes competition is by promoting specialization and differentiation among ports within the same cluster. Rather than all ports competing for the same types of cargo or market segments, clustering allows each port to focus on its unique strengths and competitive advantages. As an illustration, a specific port within a cluster might focus on managing containerized cargo, as one focuses on bulk commodities or liquid chemicals. This specialization reduces direct competition among ports within the cluster and enables them to capture a larger share of their respective markets [37].

With port data, ports will be grouped with high, medium, and low incomes. Grouping based on attributes in the form of parameters such as facility, operational, and revenue data. These attributes will be processed through Orange using a k-means algorithm to generate clusters. Testing the K-Means Algorithm on Orange, the testing process is carried out without changing the Orange app. As for the tests carried out, namely by using the k-means algorithm. Following the process is carried out in Figure 2.

2.png

Figure 2. Widget flow view on Orange application

The results of the process of testing the accuracy of the tree algorithm using the "Test and Score" in the Orange application obtained an accuracy of 95.5%, F1 Score of 95.4%, precision of 95.5%, recall of 95.5%, specificity of 87.3%, and get AUC value (Area Under Curve) of 96.1%. The AUC value is a calculated value of the performance of a classification data index. The more significant the AUC value, the better the performance index data used. A significant AUC value of 96.1% indicates that the index performance classification data includes Good Classification. The result of widget test and score can be seen in Figure 3.

3.png

Figure 3. Result of widget test and score

Based on the data mining process flow using Orange software, the data mining process in this study is the first to determine the training data on the dataset used where the training data used in this study are as many as 202 port data. Next is to enter the algorithm used in data mining processing, where using the decision tree algorithm is marked with the widget "tree". Views on the widget "tree" can be seen in the Figure 4.

4.png

Figure 4. Widget tree

5.png

Figure 5. Result of decision tree

Based on the Figure 5, from 201 classified port data with the decision tree algorithm, there are 100 ports equal, with 49.8 % having high revenue. Ports categorized as low income are 101 ports or 50.2 % of 107 ports, with increased revenue. 93.5% of them require an optimization strategy, while 6.5% require development because of insufficient facilities. Then from 101 low-income ports, 83.5% needed promotion because the facilities were sufficient, but ship visits were not significant. 11 Ports with low-income categories and inadequate facilities also require a merger with their nearest port. Because if it is developed, it will increase investment costs which may not necessarily be offset by increased income.

4.1 Clustering

The K-Means technique is a non-hierarchical grouping approach that can divide existing data into one or more categories. This approach divides data sets into groups, producing power with the qualities in 1 cluster. In contrast, Image segmentation means partitioning the images into various groups or groups that feature the same or have some similarities. K-means is a method of grouping the image partitions into groups such that at least one part of the cluster contains an image with the central area from the used part. K-Means Algorithm Process, the stages in the clustering algorithm process use the K-Means method to choose the number of clusters k. There are many ways to initialize inner cluster centres, but the primary choice at the cluster level is random. Allocate all data/objects into the nearest cluster. K-Means clustering widget design using Orange Data Mining as shown in Figure 6.

6.png

Figure 6. Widget flow view of K-mean clustering on Orange application

Figure 7 indicated the total distribution of port clusters based on their utility. Cluster 1 has three ports with a high utility category, 61 with low utility, and 5 with medium utility. Cluster 2 has four ports with high productivity, 17 medium, and 45 with low utility. Cluster 3 consists of 58 with high utility, one port is categorized as a port with low productivity, and two ports have medium productivity. If further analyzed, it can be concluded that low-productivity ports inhabit Cluster 1, cluster 2 groups ports with medium utility, and Cluster 3 is a collection of ports with high utility. The distribution of port cluster strategy can be seen in Figure 8.

7.png

Figure 7. The distribution of port clusters based on their utility

8.png

Figure 8. The distribution of port clusters based on the development strategy

The data obtained will be inputted and analyzed based on objects. Then it will be connected using the appropriate widget with what will be used and needed in this study. At the selection, stage will select the required data. The data is then adjusted as needed. Data in this research consists of 19 attributes with 201 data records. The sets of data preprocessing in this study use widgets preprocessing data by Normalizing features on Orange tools and then selecting Normalize to the interval [-1,1]. Table data is the result of displaying which attributes are set to be displayed as output in the select column widget. A series of port data processing uses the K-Means algorithm with the Orange application. Based on the K-Means Clustering results displayed in the scatter plot widget, 3 clusters are obtained. The distribution of ports in the cluster can be seen in the following Figure 9.

9.png

Figure 9. Typology according to the national port master plan

The Indonesian National Port Order is a port system responsible for operations, kinds, port structures, the National Port Master Plan, port sites, and within and intermodal connectivity and relationships with various stakeholders. National Port Plan is a national port framework that specifies port development based on socioeconomic region, geographical location, competitive edge, and environmental variables. The port has a role as the nodes in the transport network corresponding to their hierarchy, The gateway for economic activity, a place of training for changing modes of transportation, supporting industrial or trading activities, loading point of freight distribution, fabrication, and consolidation, and realizing the archipelagic outlook and state sovereignty. The Indonesian port is classified into Main port, Collector Port, and Feeder Port. Based on this classification ports in Indonesia are divided into 28 main ports, 164 collector ports, 166 regional feeder ports, and 278 local feeder ports.

However, the classification standards are arranged in a hierarchical form. Where the ports that are included in the highest hierarchy will be has a high level of each attribute/characteristic compared to other ports the hierarchy is classified as being below the port. Standard model of grouping his type of approach does have the advantage of clear and applicable standards to almost every port. However, the weakness is that it ignores characteristics unique to each port or does not open up space for grouping by using other variables/indicators. Because it may port with a hierarchy below may have higher attributes on other indicators or variables compared with ports included in the high hierarchy. Given that there is a gap in Standardization of port classification and research related to classification have not received attention which is intense in research themes around ports, to fill this gap The clustering method can be used to further group ports comprehensive and detailed. One of the advantages of the deep clustering method port grouping, namely this method can accommodate the need for grouping ports multi-dimensionally. With the advantage of being multi-dimensional like this, the indicators/variables used in grouping can be changed. In Figure 10, a clear comparison shows that the results of the existing typology are overlaid with the new typology. There is a port that has a high class according to the old port typology, but after evaluation it turns out that its utility is low and falls into the Class III category according to the Non-Commercial Port New Tipology. The distribution can be seen in Figures 10 and 11. The Classification of ports according to the new criteria shown in Table 4.

10.png

Figure 10. The distribution of ports in the cluster

11.png

Figure 11. The map of comparation existing port typology with new typology

Table 4. New port typology

In the context of port typology mapping in Indonesia, several ports can have combined characteristics depending on economic development, infrastructure and local development strategies. Therefore, the government and relevant stakeholders need to deeply understand each port's role and function to plan sustainable and coordinated development. Overall, port typology mapping in Indonesia creates a diverse picture of the role of each port in the national logistics and transportation ecosystem, from significant ports that are major trade gateways to local ports that serve remote communities. Each port impacts economic growth, distribution of goods, and inter-island connectivity. By understanding these differences, the government and stakeholders can develop appropriate strategies to increase efficiency, competitiveness and the positive impact of each port type in supporting sustainable development throughout Indonesia. The distribution of non-commercial port typologies in Indonesia includes 25% of non-commercial ports located in the western part of Indonesia, namely Sumatra Island, Java Island, West Kalimantan Province and Central Kalimantan Province. The number of non-commercial ports in central Indonesia is 42% spread across Bali, West Nusa Tenggara and East Nusa Tenggara. Meanwhile, the % of non-commercial ports in eastern Indonesia, including Maluku and Papua, is 33%.

12.png

Figure 12. Port hinterland competition

As can be seen in Figure 12, the capacity of Tanjung Emas Port has reached 855.994 TEUs, currently only around 57% is filled. The total generation in the Central Java hinterland area reached 1.3 million TEUs. Meanwhile, for non-containers, the results of the real demand survey also show that total non-container demand is still below the capacity of the ports on the north coast of Central Java. Analyzing port hinterland competition through clustering entails a comprehensive examination of various factors influencing the interplay between ports and their surrounding regions. Hinterland competition is a critical aspect of port management and maritime logistics, as it directly impacts a port's market share, operational efficiency, and overall competitiveness. By applying clustering techniques to understand hinterland competition, stakeholders can gain valuable insights into the distinct characteristics and dynamics of different port regions, thereby enabling more informed decision-making and strategic planning.

Clustering, in the context of port hinterland competition, involves grouping ports and their respective hinterlands based on shared attributes, such as geographic location, market specialization, infrastructure connectivity, and economic drivers. By categorizing ports into clusters, analysts can identify similarities and differences among competing port regions, facilitating a deeper understanding of their competitive strengths and weaknesses. This approach allows stakeholders to assess the relative attractiveness of different port hinterlands, identify emerging trends, and anticipate future challenges and opportunities.

One key aspect of clustering port hinterland competition is the recognition of regional variations in market demand, supply chain dynamics, and logistics infrastructure. Different port clusters may cater to distinct market segments, such as containerized cargo, bulk commodities, or specialized goods. For example, ports located in proximity to major manufacturing centers may specialize in handling containerized exports, while ports situated near resource-rich areas may focus on bulk cargo shipments. By clustering ports based on their market specialization, stakeholders can tailor their strategies to capitalize on unique market opportunities and competitive advantages. Furthermore, clustering enables stakeholders to assess the effectiveness of transport and logistics networks in connecting ports to their hinterlands. Ports within the same cluster may share similar transportation corridors, distribution channels, and multimodal infrastructure, influencing their accessibility and competitiveness. For instance, ports located along well-established trade routes or connected to efficient rail and road networks may have a competitive edge over ports with limited transportation connectivity. By clustering ports based on their transportation accessibility and connectivity, stakeholders can identify infrastructure gaps and investment priorities to enhance regional connectivity and competitiveness. In addition to market specialization and transportation connectivity, clustering port hinterland competition also involves considering factors such as port governance, regulatory frameworks, and institutional capacities. Ports within the same cluster may face similar regulatory environments, legal frameworks, and administrative procedures, shaping their operational flexibility and regulatory compliance. Moreover, clusters may exhibit varying degrees of institutional coordination and collaboration among port authorities, government agencies, and private sector stakeholders, influencing their ability to adjust shifting market dynamics and regulatory requirements. By clustering ports based on their governance structures and institutional capacities, stakeholders can identify opportunities for policy harmonization, regulatory reform, and collaborative initiatives to improve port competitiveness and efficiency. This holistic approach enables stakeholders to navigate the challenges and opportunities of port hinterland competition effectively, fostering sustainable growth, and prosperity in maritime trade and logistics.