Palm oil is a multi-billion dollar industry, yet to date no attempt has been made for complex knowledge modeling within the Palm Oil industry. As oil palm plantation industries contain numerous changing conditions and since relevant decision making parameters are dynamic as well, an intelligent Decision Support System (DSS) that is context sensitive, environment specific and localized for the user is needed. Our proposed solution empowers the end-users through semantic and ontology development methods, where the involvement of domain experts and end-users drive the application development.

The proposed Semantic DSS for Palm Oil is based on the idea of fusion of heterogeneous information from distributed data sources.  It will be the first ever commercially available DSS to be used in the Oil Palm Industry. Our approach is based on integrated usage of Semantic Technology through Ontologies, Context- Management and Web-services. Bayesian Network (BN) and Fuzzy Logic would be applied to the DSS to provide intelligent reasoning and to support effective decision making. In the presented work the access to the Semantic DSS is made in a unified way using Web-services. The DSS resides within the Semantic Portal for Palm Oil where the latest and most concise information is made available for the user.

1.    Introduction

A major challenge faced by oil palm plantations in Malaysia is crop productivity. This is complicated by other limitations such as labor shortage and declining planting acreage. Decision making is bogged down by several factors including the fact that the data is spread across different geographical locations and in disparate forms. Traditionally the Oil Palm plantations have been run by estate managers with vast on the job experience. Today, the oil palm industry has grown to become a multi-billion dollar industry in Malaysia, Indonesia and even in some South American countries, with top management located across the globe and managing plantations with their multi-skilled task force. They have to make crucial and timely decisions, taking into consideration factors ranging from costing, environmental issues and policies governing the operations of the palm oil industry.

 A well structured intelligent Decision Support System (DSS) can help Plantation Managers make timely agronomic decisions, perform meaningful crop monitoring and facilitate more accurate yield projections. This in turn will assist in effective expenditure control. The proposed solution facilitates end-users by allowing them to select their own relevant decision making parameters from a well-defined system that maps all their data to the ontology. It would also be possible to allow the user to include new parameters and the relationships between these parameters could then be incorporated into the reasoning power built into our DSS.

2.    Ontology and Decision Support

In AI, ontology can be defined as a formal description of concepts using basic terms and relationships as well as the rules for combining the terms in a problem domain. While abstraction of an ontology development is similar to the definition of a conceptual model, the focus is on extended definitions of relationships and concepts; and having the explicit goal of reuse and sharing knowledge through using a common framework.

Ontology development for conceptualizing knowledge components and their relationships in a formal explicit specification is not a new concept for solution developers. In the last few years, ontology approaches have become an universal technique to build explicit understandings of the structure of complex problems such as in bioinformatics, WWW design and medical informatics. In most of these cases, ontologies were used for data integration, data interoperability, and also for outlining system metadata. To date, no attempt has been made for complex knowledge modeling for the Palm Oil industry anywhere in the world.

2.1     How does the Ontology based DSS for Palm Oil Works

The proposed DSS is a wholly Semantic System. The DSS contains well managed Ontologies placed within Knowledge Bases situated dynamically within the system. The DSS is managed through SOA that allows the close integration and communication between the structured and semi-structured data in the DSS. For the DSS a multi-layered intelligent structure is proposed. Each layer’s functionality is enhanced by WSDL/ UDDI Registry and SOAP that traverse through the DSS. Through this service the data for the DSS can be located at any physical point and can be accessed by the system 24/7.

The DSS uses well researched algorithms to enable the decision making in the DSS. The purposes of the algorithms are to model and execute the parameters in a sequence that enhances effective decision making.

The DSS would be very much context-sensitive to meet the user’ requirements and another unique aspect of the DSS for Palm Oil is that the user could input his own parameters if the detailed and well established parameters and concise algorithms do not meet his requirements fully. The GUI part of the DSS is planned to receive user input and update the relevant parameters found in the Knowledge Base.

2.2     Functionality of the Multi-layered SOA in the DSS

   2.2.1    Business Service Layer

The main function of the Business Service Layer is to capture the formal and semi-formal representation of the data in the systems and databases. The Data are mainly in databases found in Palm Oil plantation offices and Oil Palm mills in Malaysia, Indonesia and elsewhere.

   2.2.2    Web Service enabled SOA Architecture

The main entities in the web service layer are: (1) the service provider that publishes the WSDL service description in the service registry in order to allow the requester to find the data required from the database or from where the data is stored. (2) The service requester that retrieves a service description directly from the database or data repository and (3) the service requester that invokes or initiates an interaction with the databases at run time using SOAP to locate, contact and retrieve (invoke) data from the databases  stored at multiple locations. The UDDI Service Registry acts as an intermediary between web service providers and requesters.

   2.2.3    Artificial Intelligence Engine

The AI Engine consists of two layers:

DSS Algorithm/ Model Layer

This layer provides the pre-conditions and effects of the decision making. The algorithm uses statistical Fuzzy Logic and heuristic actions used in decision making.

AI and Rules Component Layer

AI Component: This layer consists of the DSS Ontology and a query Engine. The DSS Ontology would contain the stored algorithm model and the rules needed to match and invoke the relevant algorithm to execute parameters to give the required response when a query is received. The DSS Ontology also reacts by giving alerts. This ability is achieved through fuzzy logic reasoning rules that are placed within the DSS. When a certain given threshold is reached or an abnormality is detected an alert is sent out to the relevant person upon login to the DSS so that a necessary action can be taken. The alert could also be mailed to the user in the form of reports.

Rules Component Layer: This Layer is made up of several  components of WinProlog; (1) Visirule, (2) Flex 4700. Both of these components work together with Bayesian Network (BN) reasoning. BN is selected to handle the uncertainty data found within the decision parameters. The combination of this trio enhances the DSS ability to handle multi varied parameters and their ratios needed in effective decision making.

   2.2.4    Semantic Layer

Several Ontologies including the Crop Upper Ontology, ENT Ontology and the User   Profile Ontology are placed here.

Crop Upper Ontology
contains concepts that are Meta, generic and abstract. The Crop Upper Ontology provides structure and general concepts upon which the Palm Oil domain ontology and other ontologies are constructed.

ENT Ontology
is the Palm Oil Domain Ontology. It contains all the concepts and variables within the palm oil industry and also the relationships between them. This ontology would integrate closely with both the User Profile Ontology and the DSS Ontology. The ENT Ontology contains the parameters or knowledge components of palm oil decision making in the form of potential classes, instances and inferred relationships.

The repository of this ontology is visible to the users through the interactive UI of the Semantic DSS. The end-users are allowed to select, create, edit or modify the parameters displayed. Updating the ontology or knowledge base according to the required changes of the user makes a ready knowledge base always. This would ensure that the decisions made are contextual to the environment of the user despite volatile, changing factors. The user would also be able to compare advice and guidance that would be provided by Domain Expert obtained from the DSS Ontology and could decide to adhere to it.

User Profile Ontology would represent all the relevant information to profile the user. Some of the information is of explicit nature, that we gather directly from the user, such as name, job designation, interest and characteristics of the user. It is important to profile the user so that accurate and timely decisions can be made by the user. The DSS is also planned to include profiling of users as they navigate through the DSS and Semantic Portal.

There is also a Semantic Annotator to annotate the information when received from structured or unstructured data.  A Query Engine is placed in this layer to handle and respond to queries received from users. The Palm Oil DSS would be able to receive queries in natural language and these queries would be understood through the NLP (Natural Language Processing) ability of the DSS.

3.0 Semantic Palm Oil Portal

“The Semantic Web is not a separate web but an extension of the current one in which information is given well defined meaning, better enabling computers and people to work in cooperation. The first steps in weaving the Semantic Web into the structure of the existing web are already under way. In the near future, these developments will usher in significant new functionality as machines become much able to process and “understand” data that they merely display at present”.

                                            Tim Berners Lee, Scientific American (May, 2001)

We are currently in the progress of designing and developing a Semantic Portal for Palm Oil. The objective of this semantic portal is to retrieve unstructured information from different web sites on the net and ‘place’ this information in an organized and structured way, that is easily accessed by users. We are applying this method for the Palm Oil industry domain, but it proves to quite general and domain independent.

Traditionally, portals are just an information centralizer that contains addresses of other web sites. Traditional portals provide a series of active links. These give access to all contents of the referenced site but not to a  specific information needed by the user. In semantic portals the portal and the referenced sites form a correspondence through explicit relationships (semantics).

When users first access our Palm Oil Portal, they would view a comprehensive website that gives them complete and latest information on palm oil. This will include research journals, articles related to palm oil industries and other valuable information in the form of natural language texts, images, videos etc. The development of the Palm Oil domain specific ontology on which the semantic portal is based on allows the annotation and classification of unstructured information available in heterogeneous formats from different web sites. When annotated, the information becomes instances of the ontology’s classes. Our portal is able to detect the navigation pattern of the user and update this information into the User Profile.  With this knowledge our Portal would recommend articles or new information that could interest the User.

Through the Web-crawler, Analyzer and Summarizer used in our Semantic Portal for Palm Oil, we are able to provide the accurate, reliable and consistent information, without overloading our user with a barrage of unnecessary facts.

 

3.0 Conclusion

Our Semantic DSS for Palm Oil and Palm Oil Portal is a work-in-progress and we expect this great project to be completed by the end of 2009. When completed the DSS becomes a very valuable tool that could help multi levels of management within the oil palm industry. They would be able to make accurate yield projections and also control the costs involved within the industry. With the inclusion of environmental factors, they would not only be making effective and efficient decisions that could help the industry but also the rest of the world.