Content area
Purpose
Fish quarantine is a measure to prevent the entry and spread of quarantine fish pests and diseases abroad and from one area to another within Indonesia's territory. Based on these backgrounds, this study aims to identify the knowledge, knowledge management (KM) processes and knowledge management system (KMS) priority needs for quarantine fish and other fishery products measures (QMFFP) and then develop a classification model and web-based decision support system (DSS) for QMFFP decisions.
Design/methodology/approach
This research methodology uses combination approaches, namely, contingency factor analysis (CFA), the cross-industry standard process for data mining (CRISP-DM) and knowledge management system development life cycle (KMSDLC). The CFA for KM solution design is performed by identifying KM processes and KMS priorities. The CRISP-DM for decision classification model is done by using a decision tree algorithm. The KMSDLC is used to develop a web-based DSS.
Findings
The highest priority requirements of KM technology for QMFFP are data mining and DSS with predictive features. The main finding of this study is to show that web-based DSS (functions and outputs) can support and accelerate QMFFP decisions by regulations and field practice needs. The DSS was developed using the CTree algorithm model, which has six main attributes and eight rules.
Originality/value
This study proposes a novel comprehensive framework for developing DSS (combination of CFA, CRISP-DM and KMSDLC), a novel classification model resulting from comparing two decision tree algorithms and a novel web-based DSS for QMFFP.
1. Introduction
The decision support system (DSS) is an implementation of the knowledge application system technology that has been widely applied in various fields (Becerra-Fernandez and Sabherwal, 2015). DSS in the field of risk assessment or risk assessment has been carried out to make decisions about the lowest risk option. Several previous studies related to DSS in the field of risk assessment, such as detection of animal/biological products, especially fish and other fishery products (Copp et al., 2016, 2021; Dowling et al., 2016; Kumschick and Richardson, 2013; Marcot et al., 2019; Trump et al., 2018).
This research will focus on the knowledge management system (KMS) of quarantine measures for fish and other fishery products (QMFFP). Risk assessment for animal quarantine measures, especially fish and fishery commodities, is very important for a country (Sampaio et al., 2015). This issue is crucial because it can support service performance in handling the import of fishery commodities as manifested in national policies and strategies. However, QMFFP in Indonesia is still inadequate. This condition is because fish quarantine actions are still carried out conventionally based on personal knowledge of fish quarantine officers, routines and instructions from the leadership or regulatory textual information. So, services for importing fish and other fishery products (FFP) cannot be carried out quickly and precisely. This conventional condition is made worse because of the rotation of fish quarantine officers. The level of knowledge of each fish quarantine officer is different, so that some officers experience difficulties in making quick and precise fish quarantine decisions by regulations. This condition results in difficulties in law enforcement for FFP importation violations. All these problems are the accumulation of the main problems. The main problems are as follows: first, the difficulty of identifying critical knowledge related to QMFFP in the form of tacit (fish quarantine officers) and explicit (QMFFP regulations). Second, the difficulty of finding new knowledge by utilizing historical FFP importation data. Third, the challenge of finding the main factors affecting quarantine for fish and other fishery products (QFFP) and the hidden relationships between them. Fourth, it is unable to determine priorities for knowledge-based system needs, so it cannot apply and utilize knowledge related to QMFFP. Fifth, there has been no effort to preserve tacit and explicit knowledge regarding QMFFP.
Recent knowledge management (KM) studies for importation and QMFFP are still very insignificant. Several studies on the importation, quarantine of live and processed commodities, especially QMFFP, focus on specific aspects, such as information technology (Hood et al., 2019; Huang et al., 2008; Mallick et al., 2020; Newton et al., 2019; Sun et al., 2017); process (Chalam et al., 2020; Hitchens and Blakeslee, 2020; Hood et al., 2019; Hood and Perera, 2016; Jang and Moffitt, 2019; Lieggi, 2020; Malhotra, 2019; Song et al., 2019); people and organizations (Hitchens and Blakeslee, 2020; Huang et al., 2008; Malhotra, 2019). The grouping of these studies shows evidence that it is minimal or even no comprehensive study in developing KM solutions (in the form of a knowledge-based system) for importation, quarantine of live and processed commodities, especially QMFFP. Comprehensive is meant in terms of a framework that can cover aspects of technology, KM processes, people and organizations. KMS is a form of KM solution that can accommodate these three aspects.
Based on all these problems, we can group them into four categories: people, technology, processes and organizations (Heisig, 2014). Human problems can be overcome with two recommended solutions: capturing the knowledge of senior QFFP officers who have taken QFFP actions according to regulations. Furthermore, developing KM technology can accelerate the exchange process (explicit knowledge transfer), for example, web-based access to data and databases. The recommended solution for technology problems is to use a knowledge discovery approach to discover new knowledge through the cross-industry standard process for data mining (CRISP-DM) methodology. It can also be pursued by creating a system of application/utilization of knowledge to support decisions (DSS). Simultaneously, the process problem can be proposed as a recommendation for analyzing the KM contingency factor table. This analysis is useful for identifying the KM process to determine the priority needs for the type of KMS in accelerating the knowledge application process. Finally, organizational problems can be faced by developing a system that can change textual regulatory information and standard operational procedures (SOPs) into ready-to-use practical information. The system is expected to accelerate the KM direction and internalization process. This study aims to answer the following research questions to address the research gap:
What are the knowledge, KM process and priority needs of KMS for fish and other fishery product quarantine measures?
What is the classification model to support decisions on quarantine measures for fish and other fishery products?
How is the DSS designed to support decisions about quarantine measures for fish and other fishery products?
This paper is structured as follows. The first part discusses the importance of DSS for QMFFP as a KM implementation, approaches to identify KMS needs and approaches to developing KMS. The following sections provide a methodology for answering RQ. Then we present the results of our study in order of RQ. In the final section, we provide a synthesis of the implications of the results for future research and discuss the study's limitations.
2. Relevant theories
2.1 Knowledge management solution
KM is a comprehensive knowledge management process to improve organizational performance through four processes, namely, discovery, capture, sharing and application (Becerra-Fernandez and Sabherwal, 2015). KM depends on two broad aspects: KM solutions, which are specific, and the foundation of KM, broader and longer-term. Since KM Solutions includes KM components (KM infrastructure, KM mechanisms, KM technology, KM processes and KM systems), to develop a KM solution, one must identify the KM process for the priority needs of the KMS. At the same time, the case of this research needs the proper KMS for the QMFFP process. Therefore, contingency factor analysis (CFA) analysis is the most appropriate method to answer these challenges because it can understand the particular circumstances in the institution and the environment related to QMFFP.
CFA is used to determine the KM Solution the organization needs. Contingency factors will affect the KM process. After the KM process is known, the KM system needed will also be identified. These results suggest that contingency factors also influence KM systems and technology indirectly. Contingency factors that influence the KM process are task characteristics, knowledge characteristics organizational characteristics and organizational environmental characteristics (Becerra-Fernandez and Sabherwal, 2015). Researchers use CFA to identify KM processes and determine priority KM system requirements related to QMFFP.
2.2 Cross-industry standard process for data mining
Significant trend in information systems is the increasing use of machine learning and data mining over the past two years (Mazaheri et al., 2020). Data mining has been widely used to develop decision support models. The CRISP-DM is a framework for translating business problems into data mining tasks and provide data mining projects independent of both the application area and the used technology (Huber et al., 2019). It is a widely adopted industry-oriented implementation of the generic Knowledge Discovery (KD) process, as described in (Huber et al., 2019). There are six phases in this CRISP-DM (Chapman et al., 2019; Purnama et al., 2020) (Figure 1). Researchers use CRISP-DM as a framework for developing a classification model for decisions related to QMFFP. The decision tree (DT) is very appropriate to use in this case because this study requires a model to predict and discover classification rules (logic) (Adnan et al., 2017).
This study uses CRISP-DM because we still think that CRISP-DM is the complete data mining methodology to meet industry needs. This assumption is based on the KDnuggets poll that CRISP-DM is the most widely used process for DM projects (Martínez-Plumed et al., 2019). Therefore, this research case is very appropriate to use CRISP-DM because this methodology is considered the de facto standard for analytics, data science and data mining projects to support decisions.
2.3 Knowledge management system development life cycle
Based on Rahman and Selwyn (2020), this life cycle (Elias M. Awad and Hassan Ghaziri Life cycle) was introduced in 2006, with significant phases of KMS development. The stages concerned with this life cycle can directly associate with dimensions in quality model. Knowledge management system development life cycle (KMSDLC) also includes activities involved in each phase.
DSS are developed to assist decision-makers in dealing with situations where there are multiple potential solutions to a problem and none is objectively better than the others; the selection of an alternative being based on the preference of the decision-makers (subjective) (Walling and Vaneeckhaute, 2020). DSS ideally has three basic, interconnected modules. The module is a database as input, a model for optimization to generate new knowledge and a user interface to generate reports and decisions (Siddiqui et al., 2018). KMS, especially DSS, is influenced by three main factors: technology, organization and environment. The most decisive organizational factor in DSS design. Because organizational structure, processes and culture greatly determine decision-making (Freier and Schumann, 2021; Miah et al., 2020). Therefore, this research requires the identification of KM processes in the organization.
The development of DSS based on data mining, machine learning and DL approaches is overgrowing. One of the most recent DSS studies using this approach is the study conducted by Nasir et al. (2021). The methodology proposed in the study was validated internally through k-fold cross-validation and then externally validated using several separate data sets (Nasir et al., 2021). Another study was conducted by Irarrázaval et al. (2021) and van Riessen et al. (2016), who developed a DT-based DSS and clustering.
Based on previous research related to DSS for the assessment of risk-importation and quarantine of fish/other fishery products, it has various methods. The research by Marcot et al. (2019) discussed DSS to identify potential invasive and detrimental freshwater fish using a quantitative approach to a Bayesian network model. Meanwhile, DSS in other studies uses fuzzy logic, artificial neural network and gray theory to produce smart hybrid algorithms to solve risk management evaluation in aquatic products for export trade in China (Wang et al., 2008). Another research related to DSS in fisheries commodities was conducted by Dowling et al. (2016) which allows users to characterize their fisheries about available data, the biological/life-history attributes of the relevant species, fisheries operational characteristics, socioeconomic characteristics and governance context. The research of Dowling et al. (2016) used a rules approach to develop its DSS. All these methods, of course, have almost the same DSS model, especially in the context of fish quarantine risk assessment. In developing the DSS model, of course, it has the same concept as CRISP-DM. Each development step has a phase of business understanding, data understanding, data processing, modeling, evaluation and deployment (Chapman et al., 2019). The lack of use of the DSS for fish quarantine measures based on the DT approach is an opportunity for new research contributions. A risk assessment DSS is needed to support quarantine measures for fish and other fishery products using the DT method.
3. Methodology
This methodology section will describe the novel framework with the specific methods and techniques used (combination of CFA, CRISP-DM and KMSDLC). The researchers present a brief research framework in Figure 2. This study uses a mixed-method approach for the following reasons. The first justification is because this study explores various factors in depth from the perspective of participants/informants directly in the actual environment. Furthermore, the obtained factors are measured by description and the relationship between factors (including predicting) with the quantitative method by collecting specific object data and meeting the statistical rules to generalize the object population. Finally, the relationship/prediction results between these factors are then validated/confirmed or further deepened using the qualitative method (Saunders et al., 2019).
3.1 Contingency factor analysis
This part of the CFA aims to develop a KM Solution design, which consists of the following stages (Becerra-Fernandez and Sabherwal, 2015). This contingency stage should include all possibilities and will work with appropriate methods. Several categories of contingency factors that influence the selection of the KM process are task characteristics, knowledge characteristics, organizational characteristics and environmental characteristics (Becerra-Fernandez and Sabherwal, 2015; Humani et al., 2020; Irawan and Samsuryadi, 2019). Based on these factors, the CFA stage is very appropriate to identify the priority needs of the organization (Becerra-Fernandez and Sabherwal, 2015; Sensuse et al., 2020).
First, analyzing contingency factors. At this stage, the contingency factors affecting the QFFP agency will be analyzed through interviews with five expert sources related to QFFP. The resource persons have one person with a doctoral education background in fisheries, three people from master's degree in fisheries and one person from bachelor's degree in fisheries. Resource persons have various positions, namely, as structural and functional officials related to QFFP. Contingency factors that influence the KM process have been discussed in the previous section. Next, identify the KM process. Based on each contingency factor, the KM process that supports the contingency factor will be selected at this stage. Third, prioritize the KM process. At this stage, scoring will be carried out for each KM of the contingency factor process. A value of 1.0 will be given if the KM process supports the contingency factor and 0.0 if not. A value of 0.5 will be given if the KM process supports each contingency factor. Then identify the KM processes that have been applied previously. At this stage, identification of the KM processes in the company will be carried out. Finally, identify additional KM processes and KMS priority. After getting the results from stages 3 and 4, the KM process obtained in stage 4 and stage 3 will be compared. If there are differences, it will be identified whether the KM process needs to be added.
3.2 Cross-industry standard process for data mining
This section aims to develop a classification model for QMFFP, which consists of the following phases (Figure 3). Figure 3 shows the 5 phases of CRISP-DM for developing the model. Figure 3 also clearly shows that this study used a mixed-method approach.
The C5.0 algorithm is among the best algorithms for calculating the separation based on the information gain ratio (IGR) (Kuhn and Johnson, 2013). IGR is a probability-based measure used to calculate the degree of uncertainty reduction. The C5.0 algorithm procedure consists of the following steps:
initialization of training sample weights;
obtaining a subset of training sequentially;
calculating subset errors and updating weights; and
evaluation of classification results (Guo et al., 2021).
Learning with the CTree algorithm was chosen because it is a comparison and is by the case of this study (Hothorn et al., 2015). The suitability of this case study can be demonstrated by the ability of CTree to identify homogeneous subgroups from within the initial heterogeneous population. In addition, CTree can perform a thorough search of all possible splits of the QMFFP variable input and select the covariates that show the best separation (Ferré et al., 2019).
3.3 Knowledge management system development life cycle
This section aims to develop a DSS related to QMFFP. KMSLDC has eight stages. The first is an evaluation of the availability of infrastructure. This stage consists of 3 steps, and the first step is to conduct interviews with three experts. The resource person has an educational background of one master in a computer, one master of information technology and one bachelor of fisheries. The second step is to conduct an aligning knowledge and business strategy (Tiwana, 2000). Third, evaluate Hardware, Software, Brainware and Processes using context diagrams.
The second stage is to form a KM team. At this stage, the researchers formed a team with a delegation of leadership and agency staff contributions. At this stage, the core step is to identify the key stakeholders of the prospective KM system based on the team members' capabilities, team size and project complexity. The third stage is capturing knowledge. At this stage, the researcher captures knowledge through interviews, special meetings, observation and classification models: interviews, special meetings and observations to capture tacit knowledge. At the same time, the classification model is used to capture and create new explicit knowledge. This process is carried out because knowledge capture and knowledge creation are critical factors for decision-making (Agrawal, 2020). The fourth stage is the DSS design. At this stage, the researcher designed the DSS architecture using the KMS architecture layer. Next is designing the software specifications. Finally, designing a web-based database. The fifth stage is DSS testing. At this stage, the researchers conducted verification and validation of the DSS, which was only limited to the predictive function and general features of the QMFFP DSS. The verification procedure aims to ensure that the system is functioning correctly. The verification procedure aims to ensure that the system has the correct output. The sixth stage is the implementation of the QMFFP DSS. At this stage, the researchers conducted training and testing the DSS to users. The seventh stage is managing the rewards change, and structure. This stage aims to support the development of the QMFFP DSS functionality. At this stage, the researcher provides several recommendations for managing changes in knowledge and organization. The last stage is the postsystem evaluation. At this stage, the researcher assesses the DSS's impact in terms of the effect on people, procedures and business performance related to the importation of QFFP.
4. Results
4.1 KM process and priority needs for KMS
The first stage is the analysis of the contingency factors. The results of the interviews at this stage are shown in Table 1. Table 1 also shows four characteristics as contingency factors. The characteristics in the table are determined based on the majority of the characteristics of the division.
The second stage is to assess KM process priorities. At this stage, we will weigh the contingency factors to assess the priority of the KM process. If the KM process supports the current contingency process, then based on Fernandez, a Yes value will be given, and the multiplier will be multiplied by 1. If No is given and the multiplier factor is 0, whereas if the process supports both types of contingency factors, an OK score and a factor will be given, the multiplier is 0.5. Based on the score obtained by each KM process, the priority needs for the KM process can be generated. The cumulative score column calculates the cumulative score of each process. The presentation is calculated based on the cumulative score and the maximum score for each process. The ranking column is a priority for the KM process. It can be concluded that the most needed KM processes based on the order are Combination, Direction, Routines, Exchange, Socialization for Knowledge Discovery, Externalization, Internalization and Socialization for Knowledge Sharing. Then the researchers did a mapping of the KM process. These results will map the needs of the KM process and the current KM process (Table 2).
Based on Table 3, the ranking column is a priority for the KM process. It can be concluded that the most needed KM processes based on the order are Combination, Direction, Routines, Exchange, Socialization for Knowledge Discovery, Externalization, Internalization and Socialization for Knowledge Sharing. Then the researchers did a mapping of the KM process. These results will map the needs of the KM process and the current KM process in Table 4.
The final stage is to analyze and recommend the priority needs for KM technology and KMS features to be developed. The results of this last stage are shown in Table 5. Table 5 shows that the highest priority requirement for KM technology is data mining with the KMS feature of the QMFFP classification knowledge model. The KMS developed focuses on improving workplace safety and QMFFP operational efficiency (Tsang et al., 2018).
4.2 DT classification model
In this first phase, the development of a decision classification model related to QMFFP consists of the following results. The results of the interview analysis in this first part consist of three essential points. First, data mining aims to determine the main factors that affect QMFFP and the relationship of causality between them. Table 6 shows the factors and target variables for QMFFP.
The second interview results in the first phase stated that the QMFFP process in the information system was application, isolation, detention, rejection, release, destruction and return. Business processes can be done through a web-based information system but have not developed a DSS. Furthermore, the data mining project boundaries that will be undertaken are fish, carrier media, SOPs, violations and other provisions stated in the regulations. In this second phase, the first interview results show that the QMFFP data has a validity that can be accounted for because it is stored in a database that has been validated by QFFP supervisors and officers. Historical QFFP data has not been used to support QMFFP decisions from the utilization aspect, knowing hidden patterns between QMFFP factors. This utilization can be done using a knowledge discovery approach with statistical methods or data mining to develop a decision model. Furthermore, the decision model can be applied to a knowledge-based. Next, collect data into one dataset table. The table consists of 22 attributes/columns. Twenty-one as the factor/predictor and one as the target class. Meanwhile, from the quality aspect, there are still a lot of blank data. Therefore, it is necessary to preprocess the data before learning by the DT algorithm. Finally, the results of the exploration of the dataset are to hypothesize whether all factors are the main factors affecting QMFFP? Furthermore, how are the related factors?
The data's preprocessing phase produces a dataset that is ready for learning processing with a DT algorithm. The dataset has 10,113 historical QMFFP records. The datasets number is vast and of high quality because the data exported from a daily transactions database has been validated. Researchers process the dataset using RStudio as an application for statistical computing and visualization. This condition is an added value to carrying out the knowledge discovery process because high-quality datasets and visualization tools for pattern exploration support discovering new knowledge (Miah and Vu, 2020). At this stage, the researcher has divided the dataset into two training data and test data.
The fourth phase produces the C5.0 and CTree DT models. Both of these models have perfect accuracy (100%). Both of these models have perfect accuracy (100%), as shown in Table 7. The difference between these two models is the execution time, the number of selected attributes, the DT and the number of rules, as shown in Figures 4 and 5. The C5.0 DT model has several attributes of 5 while CTree 6. The C5.0 DT is more straightforward than the CTree. The DT model C5.0 has the number of rules 7 while CTree 8. Based on these results, it can be concluded that the longer the execution time, the more attributes and rules will be produced.
The fourth phase's last stage is to obtain the learning test results using a confusion matrix (CM) table (Figure 6). Both models have the same CM table because they have 100% accuracy. In performance measurement using confusion matrix, there are four terms representing the results of the classification process. The four terms are true positive (TP), true negative (TN), false positive (FP) and false negative (FN). TN value is the number of negative data detected correctly, while FP is negative data but detected as positive data. Meanwhile, TP is positive data that is detected correctly. FN is the opposite of true positive, so the data is positive, but is detected as negative data. Based on CM test data, it can be concluded that all were classified correctly or the accuracy is perfect 100%. This is because there is no FALSE value in the table.
The Ctree model was chosen because it involves the QFP Status (Q) factor, which is the main factor for quarantine, especially live fish. The CTree model has six main attributes/factors that influence QMFFP. Other factors influence the main factor. Therefore the auto-fill feature is needed when developing the interface system later.
4.3 Decision support system
The first stage of evaluating the availability of infrastructure has three main results. Table 8 shows the results of the interview analysis as the first result of this stage. The analysis shows that each interview's results indicate the condition of the availability of KM infrastructure. Figure 7 shows the second result (aligning knowledge and business strategy) (Tiwana, 2000). Figure 7 also shows that the recommended KM technology is data mining, building DSS, web-based access to data and databases. The third result is KMS can be developed based on feasibility (economic, technical and behavioral). This result uses a context diagram to determine the needs and evaluation of existing hardware, software, process and brainware infrastructure. Based on the context diagram in Figure 8, the software needed is a web-based application. Next is the need for a web server with a specification of PHP version> 7 and MySQL version database> 5. Based on observations, the software needs are by the needs because it has competent human resources in server management. The required web application must have the main features of QMFFP prediction and the QMFFP database. General users can fill in the QMFFP factor to get a QMFFP prediction. Meanwhile, the administrator can access the user's input data so that all predictions made by the user will be stored in the database through the user input data menu. Based on the context diagram, the process is very likely to be developed, and by the QFFP administration process conditions. Finally, the hardware evaluation results show that the current hardware availability is by the software requirements specifications.
In the second stage, the researchers obtained the details of the KM team with the following details. First are the main stakeholders, namely: QFFP import applicants, fish quarantine officers, related agency units QFFP actions: subsector for the prosecution of QFFP violations and subsector of risk management, Fish Quarantine and Quality Control Agency/Badan Karantina Ikan dan Pengendalian Mutu (BKIPM), fish disease experts. Second, the team's capabilities consist of risk assessment, fish diseases, QFFP import SOP, system analyst, code igniter framework-based PHP (Hypertext Preprocessor) web programmer, IT infrastructure and laws related to QFFP importation violations. Third, in this case, the team's size is a joint team from several BKIPM agency units totaling 5–10 coordinators who have various competencies, namely, IT, QFFP policies and fish diseases. Fourth, project complexity is medium because it requires a team with declarative and procedural knowledge related to QMFFP, data mining, DT algorithms and web programming.
In the third stage, the researcher obtained tacit knowledge through interviews and explicit (CTree classification model) described in the previous section. The knowledge results at this stage have already been validated in the previous section. Therefore, in the third stage, KMSDLC will refer to or use the previous section results. In the fourth stage, the researcher obtained results in three essential points. First, the DSS architecture that has been designed uses the KMS architecture layer (Figure 9). Figure 9 shows a DSS architecture consisting of seven layers. The seventh layer is the layer that has the highest complexity because it contains data mining. The first to fourth layers are the user-accessible DSS web application layers. The four layers use the model-view-controller (MVC) application concept. MVC is an application that separates data (Model) from the view (View) and how to process it (Controller). The rules of the classification model results are stored at the seventh layer in the form of a database, while in the fourth layer, the controller functions are in the form of PHP programming code. The database design consists of 23 tables, 22 factors tables that affect QMFFP, 1 general function table for user authentication and 1 table for the final action target variable.
The result of this fifth stage is a DSS that has been tested based on verification and validation procedures. The verification of the DSS function has been able to provide a decision on fish quarantine measures. Figure 10 shows the interface on the main features of a DSS. Meanwhile, general system functions such as login, filter, search and data input can be done. However, some functions of updating data and filling automatic forms cannot be done. Validation of the predictive output of fish quarantine action decisions is by stakeholder regulations and evaluations. The system's general function output validation is mostly correct, but some links on the administrator page do not point to the correct page. The result of this sixth stage is the user response. Users do not experience significant difficulties when using, and there are only a few uncomfortable pages because they are not responsive. Some administrator users try that the edit page cannot be accessed because the application is still limited to storing QMFFP data. An explanation of the QFFP action solution is required after the user clicks submit. A search menu is required in the section of public pages regarding carrier types, illustrations of valid/invalid document types, etc. Other factors influence the main factor of model classification results. Therefore, we need an automatic form filling feature when filling out forms. In the seventh stage, we get the results in recommendations for managing the changes in knowledge and organization. The recommendation is that knowledge changes (in the form of rules) will continue to be updated when new cases arise by saving the QMFFP history. Second, it is recommended that the Communities of Practices (CoP) method be used to discuss the development of changes related to QMFFP. CoP's role is to explore how individuals work, share knowledge and improve operational practices (Jassbi et al., 2015; Jørgensen et al., 2020; Lanke and Nath, 2021). Third, a reward mechanism is recommended for fish quarantine officers who share knowledge or ideas regarding new QMFFP factors and features for developing DSS functionality.
The final stage the researcher obtains is an evaluation of the system and the overall impact on people, procedures and business performance regarding the importation of QFFP. Table 9 shows the results of this stage. Table 9 also shows the impact and evaluation of DSS by the dimension of KM.
5. Conclusion and future works
This research has been able to answer three research questions. RQ1's answer is as follows. The need for priority knowledge that is not yet possessed is declarative knowledge in knowing the main factors of fish quarantine action and its relationship. The KM processes that are most needed based on the sequence are Combination, Direction, Routines, Exchange, Socialization for Knowledge Discovery, Externalization, Internalization and Socialization for Knowledge Sharing. Priority KMS are the knowledge discovery system (KDS) and knowledge application system (KAS). The highest priority requirement for KM technology is data mining and DSS. The answer to RQ2 states that the best decision model most appropriate for the QMFFP case is a model with the CTree algorithm. At the same time, the responses to RQ3 indicate that a web-based DSS can speed up QMFFP decision-making. Based on these results, the development of KMS, especially in the form of a DSS for QMFFP, is strongly affected by several aspects related to managing FFP. These aspects are knowledge type, KM processes, priority needs regarding business processes, KM technology, stakeholder environment and organizational structure and culture. All of these answers indicate that the research results have been able to achieve the objectives of this study. The three main objectives of the research (priority of KMS requirements, data mining models and DSS) have been achieved using the combined methods set out in the research framework. The contingency factor method is proven to be able to identify priority KMS needs. CRISP-DM and DT algorithms as data mining methods can develop a more effective QMFFP classification model. KMSDLC as a framework can develop a more systematic and planned KMS.
The implication of the development and implementation of this KMS is that there are significant changes to management practices of human resources, processes, service products, policies and organizational performance related to QMFFP. Human resources become easier to learn new knowledge, adapt, rotate positions and achieve job satisfaction. The decision-making process for QMFFP becomes more high quality, effective, efficient and innovative. QMFFP public service products have added value, easy access and are more automated because they are already in the form of a knowledge-based online system. The policy on QMFFP is further enhanced by the existence of knowledge-based service performance standards, rewards for sharing knowledge and knowledge preservation procedures. Organizational performance becomes more adaptive to service users' conditions to be more loyal in following organizational procedures. Furthermore, the organization will focus more on developing resources and knowledge (tacit and explicit).
This research also contributes to management theory. The contribution is to propose a framework as a solution for strategic management. The framework consists of identifying knowledge-based system requirements, a classification model based on data mining, and DSS development as a knowledge product. This framework will enrich management theory in providing appropriate technology-based strategy recommendations to support decision-making.
This study has limitations that can be used as input for further research. First, it is still necessary to add other factors in the future related to specific laboratory actions. Furthermore, QFP experts provide advice on detailing QFP options on QFP attributes. Finally, this study still manually updates the knowledge/rules into the DSS. The future research opportunity is to update knowledge/rules in real-time when there are new cases/factors. Besides, future research can perform data mining using other methods. Future research has the potential to use artificial intelligence approaches to develop smarter models to support decisions. Examples include deep learning models, knowledge graphs and natural language processing. In addition, the development of the QMFFP knowledge discovery framework is an excellent opportunity to be studied in the future as a recommendation for fish quarantine agencies.
Meanwhile, this research's solution is still minimal on increasing the sharing culture from the aspect of impact. Therefore, further research is needed to develop other types of KMS, such as knowledge sharing systems and knowledge capture systems. In addition, the opportunity to examine the knowledge-sharing model is wide open. This model is expected to identify the components that influence and impact the knowledge-sharing culture.
This study also has limitations in determining experts and identifying critical knowledge for QMFFP. Expert assessment should use criteria and methods. Identification of critical knowledge ideally uses knowledge mapping, analytical hierarchy process and knowledge loss risk matrix approaches. Knowledge mapping about QMFFP is a very potential topic to be done in the future. Research on this topic is urgently needed to visualize the sources and flows of QMFFP knowledge (tacit and explicit) (Anthony, 2021). Furthermore, this topic can design a knowledge mapping model to provide a blueprint for a knowledge mapping-based system. Develop a system that provides visualization of operational information to facilitate interaction and communication between quarantine officers, laboratory officers, fish disease researchers, applicants and other stakeholders.
The authors would like to thank Direktorat Riset dan Pengabdian Masyarakat (DRPM) from University of Indonesia for funding this research through the “Publikasi Terindeks Internasional (PUTI) Q2 Tahun Anggaran 2020 Nomor: NKB1479/UN2.RST/HKP.05.00/2020” program. The first authors are the main contributors to this paper.
CRISP-DM phases
Research framework
CRISP-DM phase for developing the model
Decision tree model with the C5.0 algorithm
Decision tree model with the CTree algorithm
Confusion matrix (CM) table. (a) C5.0, (b) CTree
Strategic knowledge gap analysis
DSS context diagram
DSS architecture
DSS interface
Analysis of contingency factors
| Task Characteristics | |
|---|---|
| Task Uncertainty | Low |
| Task Interdependence | High |
| Knowledge Characteristics | |
| Tacit vs Explicit | Explicit |
| Declarative vs Procedural | Procedural |
| Organizational Characteristics | |
| Organizational Size | Large |
| Business strategy | Differentiation |
| Environmental Characteristics | |
| Environmental Uncertainty | High |
Analysis of KM process needs
| Contingency |
Combination | Socialization for |
Socialization for |
Exchange | Externalization | Internalization | Direction | Routines |
|---|---|---|---|---|---|---|---|---|
| Environmental uncertainty | Y | Y | N | N | N | N | Y | Y |
| Business strategy (differentiation or low cost) | Y | Y | N | N | N | N | N | N |
| Organizational size | Ok | N | N | N | Y | Ok | N | Y |
| Procedural/declarative | N | N | N | N | N | N | Y | Y |
| Explicit or Tacit Knoweldge | Y | N | N | Y | N | N | Y | N |
| Task interdependence | Y | Y | Y | Y | N | N | Y | N |
| Task uncertainty | Y | N | N | Y | Y | Y | N | Y |
| "YES" | 5 | 3 | 1 | 3 | 2 | 1 | 4 | 4 |
| "NO" | 1 | 4 | 6 | 4 | 5 | 5 | 3 | 3 |
| "OK" | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
| Score | 5.5 | 3 | 1 | 3 | 2 | 1.5 | 4 | 4 |
Analysis of KM process needs
| Contingency Factors KM Process | Cumulative Score | Maximum Score | (%) | Grade |
|---|---|---|---|---|
| Combination | 5.5 | 6 | 91.66667 | 1 |
| Direction | 4 | 6 | 66.66667 | 2 |
| Routines | 4 | 5.5 | 66.66667 | 3 |
| Exchange | 3 | 5.5 | 50 | 4 |
| Socialization for Knowledge Discovery | 3 | 6 | 50 | 5 |
| Externalization | 2 | 6 | 33.33333 | 6 |
| Internalization | 1.5 | 6 | 25 | 7 |
| Socialization for Knowledge Sharing | 1 | 6.5 | 16.66667 | 8 |
KM process needs mapping
| KM Activity | Contingency Factors KM Process | KM Process | Current |
Priority |
|---|---|---|---|---|
| Discovery | Combination | High | High | 1 |
| Socialization for knowledge discovery | High | High | 3 | |
| Capture | Externalization | Low | Low | 2 |
| Internalization | Low | Low | 4 | |
| Sharing | Socialization for knowledge sharing | Low | High | 3 |
| Exchange | High | Low | 2 | |
| Application | Direction | High | High | 1 |
| Routines | High | High | 1 |
KM process needs mapping
| Priority | Contingency Factors KM Process | Needs | KM Technologies | KMS features |
|---|---|---|---|---|
| 1 | Combination | Main factors and classification models for QMFFP | Data mining | QMFFP classification model |
| 1 | Direction | Supporting QMFFP action decisions | Decision support systems | Prediction of QMFFP |
| 1 | Exchange | Dissemination of QMFFP case history data | Web-based access to data, databases | QMFFP database |
| 2 | Routines | Public users, fish quarantine officers and administrators in administration and reports on QMFFP | Management information systems | CRUD management and QMFFP reports |
| 2 | Externalization | Best practices and action cases for QMFFP | Best practices, lessons learned databases | A database of cases and factors of QMFFP action |
| 3 | Socialization for knowledge discovery | QMFFP CoP discussion | Electronic discussion groups | Forum |
| 3 | Socialization for knowledge sharing | QMFFP CoP discussion | Electronic discussion groups | Forum |
| 4 | Internalization | Factor learning and QMFFP action simulation | Simulation of the import flow of QMFFP | QMFFP History database |
Factors/predictor and target variables for QMFFP
| No. | Code | Description | Variable type |
|---|---|---|---|
| 1 | CM | Name of fish/ carrier media | Factors/predictor variables |
| 2 | A | Types of fish (ornamental fish, consumption fish, etc.) | Factors/predictor variables |
| 3 | B | Class (pisces, amphibians, or others) | Factors/predictor variables |
| 4 | C | Group (life and dead/ specimen) | Factors/predictor variables |
| 5 | D | Form/type/processing (dry, processed, fresh, etc.) | Factors/predictor variables |
| 6 | E | Acquisition of carrier media (cultivation or industrial products) | Factors/predictor variables |
| 7 | F | Intended use (hobbies, consumption, feed) | Factors/predictor variables |
| 8 | G | Registration status (not/not registered or already registered) | Factors/predictor variables |
| 9 | H | Susceptible species (yes or no) | Factors/predictor variables |
| 10 | I | Infected*/ Non OIE (yes or no) | Factors/predictor variables |
| 11 | J | Level of risk category (high or low) | Factors/predictor variables |
| 12 | K | The carrier is managed or known to the owner (yes or no) | Factors/predictor variables |
| 13 | L | Through designated entry points (yes or no) | Factors/predictor variables |
| 14 | M | Apply for import by the owner/proxy through the fish quarantine online PPK and report to the quarantine officer (yes or no) | Factors/predictor variables |
| 15 | N | Document status (1. valid [document no 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11], 2. invalid [document no. 4, 5, 6, 7], 3. invalid [document no 1, 2, 3, 8, 9, 10]) | Factors/predictor variables |
| 16 | O | Suitability of the content (type, quantity and/or size) of the carrier media with the accompanying documents (1. appropriate, 2. inappropriate/rotten/damaged) | Factors/predictor variables |
| 17 | P | Isolation in fish quarantine installation (yes or no) | Factors/predictor variables |
| 18 | Q | Quarantine fish pest (QFP) status (1. QFP was not found after isolation at the fish quarantine installation, 2. QFP was not found without isolation due to low risk, 3. QFP was found that did not allow treatment, 4. QFP group I and group II were found) | Factors/predictor variables |
| 19 | R | Detention (1. Yes, 2. Not detained because it meets the requirements, 3. Not being detained because it has the potential to be rejected) | Factors/predictor variables |
| 20 | S | Fulfill the shortage of requirements (1. meet the requirements after being detained (before 3 [three] days), 2. meet the requirements without detention, 3. fail to meet the requirements) | Factors/predictor variables |
| 21 | T | Rejection (1. Yes, 2. Not Rejected because it is by the requirements, 3. Not Rejected because it has the potential to be destroyed) | Factors/predictor variables |
| 22 | V | Final QMFFP {1. Importation exemption because it meets the requirements; 2. Delivery to the country of origin because the applicant agrees (sent 3 [three] days after being rejected); 3. Destruction for Disagreeing/Not willing to be sent to the country of origin after 3 (three) days of refusal; 4. Extermination due to damaged/rotten MP; 5. Destruction because QFP groups I and II were found.} | Target variables |
Notes:PPK = Permohonan pemeriksaan karantina; OIE = Office International des Epizooties
The results of the decision tree classification model
| Data Uji | 60:40:00 | 65:35:00 | 70:30:00 | 75:25:00 | 80:20:00 |
|---|---|---|---|---|---|
| Accuracy (%) C5 and Ctree | 100 | 100 | 100 | 100 | 100 |
| Number of rules C5 | 7 | 7 | 7 | 7 | 7 |
| Number of rules CTree | 8 | 8 | 8 | 8 | 8 |
Analysis of interview results to define the availability of KM infrastructure
| No. | The interview result (the current conditions) | Analysis and Evaluation | KM infrastructure |
|---|---|---|---|
| 1 | The use of technology is still minimum, still limited to information systems | There are still obstacles and opportunities to improve importation and QMFFP services. Examples are QFFP officer issues, process problems and policy issues. Meanwhile, the opportunity is to provide tools to simplify QFFP services | This analysis shows that the organization has an IT infrastructure |
| 2 | The most influencing factor is the rotation of QFFP officers, mentorship culture and group communication, which causes different QMFFP data retrieval | The subjectivity of each employee is because each officer has different knowledge regarding QFFP. Therefore, the organization needs a system that can support decisions without depending on the rotation of officers | This analysis shows that the organization has an organizational culture, communities of practice and common knowledge |
| 3 | The QMFFP process is very complicated | This condition encourages the need to prioritize KM processes for organization structure. This effort aims to meet the demands of the smooth flow of the process of entry and exit of fishery commodities, which frequency is getting higher | This analysis shows that the organization performs KM processes and organizational structure (though minimal) |
| 4 | Decision-making for QMFFP is still based on the experience and tacit knowledge of officers | The knowledge required is both declarative and procedural knowledge, including QMFFP factors and basic knowledge of the steps to be taken to decide QMFFP | This analysis shows that the organization has an organizational culture and common knowledge |
| 5 | KM technology has not been applied for knowledge discovery and knowledge application related to QMFFP | Develop KM technologies such as data mining and DSS for QMFFP | This analysis shows that organizations carry out a process of discovery and application of knowledge |
| 6 | Implementation of regulations and sources of knowledge in QMFFP is still conventional | Even though it already has an information system but is not knowledge-based. Therefore, it is necessary to develop a knowledge-based system for QMFFP that stakeholders can access | This analysis shows that the organization has an IT infrastructure |
| 7 | Efforts to manage and preserve knowledge related to QMFFP already exist but are not yet structured | Storage and provide QMFFP knowledge retrieval feature in a web-based database | This analysis shows that the organization has started to organize knowledge |
Impact and evaluation of DSS
| No. | Dimension | Impact | Evaluation |
|---|---|---|---|
| 1 | People |
|
The data input feature in the user interface must beequipped with references and illustrations, for example illustrations and tutorials to select valid documents |
| 2 | Processes |
|
|
| 3 | Product |
|
|
| 4 | Organizational performance | Quarantine institutions have a good reputation for service, but the socialization of the QMFFP program has not yet been carried out massively |
|
Note:TAM = Technology acceptance model; TOE = Technology organization environment
© Emerald Publishing Limited.