An Intelligent Recommender E-Business for Export Oriented Coconut Industry Based on Web Mining and Radar Chart Zafira Kanara Email: z_kanara@yahoo.com Dr. Ir. Yandra Arkeman, M.Eng Email: yandra_ipb@yahoo.com Dr. Eng. Taufik Djatna,S.TP, M.Si. Email: taufikdjatna@ipb.ac.id Agroindustrial Technology Department, Faculty of Agricultural Technology, Bogor Agricultural University, IPB, Darmaga Campus, PO BOX 220 Bogor, West Java, Indonesia. ABSTRACT E-business, such as coconut based industry, has changed the whole outlook of traditional trading behavior especially for export oriented. However, the massive product information provided by the Internet Merchants causes the problem of information overload and this will reduces the customer’s satisfaction and interest. To overcome this problem, a recommender system based on web mining and radar chart is proposed in this paper. In this study, clustering using K-Means method is used. Radar chart uses graphic displayed in a Web form to evaluate multiple alternatives based on multiple criteria. This work helps to categorize and evaluate the product quality of customer preference. The results of this study show that the proposed system is potentially to give sensible recommendations, and be able to help customers make decisions. This research aims to design a web based intelligent business systems (intelligent e-business) for coconut commodities through developing an online transaction system, recommendation, and customer services. Scope of this research was a system designed internet business system for coconut commodities that include to give recommendation, customer service, products information, sale and purchase transactions. This website was built using Adobe Dreamweaver CS4 (Adobe, 2008), Adobe Photoshop CS3 (Adobe, 2007), MySQL (Oracle, 2009), and Sybase Power Designer 15.3 (Sybase, 2010). In the package of programs, there are two interface. First is the appearance of the user interface that can be accessed by customers and prospective customers or commonly called user front-end. The second view is a display interface exclusively accessed by the administrator or commonly called the user back-end. Keywords: Clustering, Radar chart, Export oriented coconut trade, E-business

I.

INTRODUCTION

recommendation system shows the importance of what customers need and help them what to buy. II. RELATED WORK This research is related to previous work conducted by several researchers. (Zeng 2009; Wanurmarahayu 2011; Ahn 2006). Zeng (2009) wrote about Intelligent E-Commerce Recommender System Based on Web Mining. This research system utilizes web mining techniques to trace the customer’s shopping behavior and learn his/her up-to-date preferences adaptively. The experiments have been conducted to evaluate its recommender quality and the results show that the system can give sensible recommendations, and is able to help customers save enormous time for Internet shopping. The research related to clustering conduct by Ahn ( 2006). The research in this paper applied K-means clustering whose initial seeds are optimized by GA, which is called GA K-means, to a real-world online shopping market segmentation case. This study compared the results of GA Kmeans to those of a simple K-means algorithm and selforganizing maps (SOM). Another research related to e-commerce conduct by Wanurmarahayu (2011) with the title Design of Internet Business System (E-Business) for Leather Agroindustry. This research aims to design a web based business system (ebusiness) for leather commodity, through developing an online transaction system (e-commerce) and customer service.

Nowadays, the advance of Internet and Web technologies has continuously boosted e-business, such as coconut based industry. Through the Internet application, coconut seller and customers can now easily interact each other, and then make transactions within a specified time. However, the Internet infrastructure is not the only decisive factor to guarantee a successful business in the electronic market. With the continuous development of electronic business, it is not easy for customers to select merchants and find the most suitable products when they are confronted with the massive product information in Internet. In the whole shopping process, customers still spend much time to visit a flooding of retail shops on Web sites, and gather valuable information by themselves. This process is much timeconsuming, even sometimes the contents of Web document that customers browse are nothing to do with those that they need indeed. So this will inevitably influences customers’ confidence and interests for shopping in Internet. In order to provide decision support for customers, one way to overcome the above problem is to develop intelligent recommendation systems based on web mining and radar chart. A system of mining method to be used in this study is clustering using K-Means method. K-Means method can classify the product by its dissimilarity using Euclidean distance. Radar chart uses graphic displayed in a Web form to evaluate multiple alternatives based on multiple criteria. Radar chart can evaluate the product quality of the product and the web form shows the size of the gaps among five to ten organizational performance areas. In the shopping websites, the system can help customers find the most suitable products that they would like to buy by classifying product and showing radar chart for each product. Therefore, intelligent e-business in

III. METHODS 1.Clustering K-Means Methods The word of a cluster which implies a bunch of things of the same kind or a group of similar things is

becoming popular now in a variety of scientific fields. This word has different technical meanings in different disciplines, but the study in this research is data clustering using KMeans method (Figure 1). Thus clustering is a technique to generate groups of subsets of data in which a group called cluster is dense in the sense that a distance within a group is small, whereas a distance between clusters is sparse in that two objects from different clusters are distant. This vague statement is made clear in the formulation of a method. First, a set of objects to be clustered is given. An object set is denoted by X  {x1 ,..., xN } in which

(iii) [triangular inequality] m(x, y) ≤ m(x, z) + m(z, y). Remark that a metric can be used as dissimilarity, while a dissimilarity need not be a metric, that is, the axiom (iii) is unnecessary. A typical example is the Euclidean metric:

d 2 ( x, y ) 

 (x j 1

p

j

 yj)  x y
2

2

where

x   x1 ,..., x p  and y   y1 ,..., y p  are p-dimensional

xk ,  k  1, 2,..., N  is an object. With a few exceptions, p p

R p . We sometimes use the Euclidean norm x 2 and the Euclidean scalar product denoted by vectors of

x1 ,..., xN are vectors of real p-dimensional space R . A generic element x  R

( x, y )   xi y j (Sadaaki Miyamoto 2008). i 1

p

x1 ,..., x p ; we write x   x1 ,..., x p   R p .

is the vector with real components 2. Radar Chart The radar chart is a chart and/ or plot that consists of a sequence of equi-angular spokes, called radii, with each spoke representing one of the variables. The data length of a spoke is proportional to the magnitude of the variable for the data point relative to the maximum magnitude of the variable across all data points. A line is drawn connecting the data values for each spoke. This gives the plot a star-like appearance and the origin of the name of this plot. The star plot can be used to answer the following questions (NIST/SEMATECH 2003): (i) What variables are dominant for a given observation?

(ii) Which observations are most similar, i.e., are there clusters of observations? (iii) Are there outliers? Radar charts are a useful way to display multivariate observations with an arbitrary number of variables. Each observation is represented as a star-shaped figure with one ray for each variable. For a given observation, the length of each ray is made proportional to the size of that variable. Radar charts differ from glyph plots in that all variables are used to construct the plotted star figure. There is no separation into foreground and background variables. Instead, the star-shaped figures are usually arranged in a rectangular array on the page. It is somewhat easier to see patterns in the data if the observations are arranged in some non-arbitrary order, and if the variables are assigned to the rays of the star in some meaningful order (Friendly 1991). Each star represents a single observation. Typically, radar charts are generated in a multi-plot format with many stars on each page and each star representing one observation (NIST/SEMATECH 2003). These are the steps to develop radar chart (Rogers 1995): 1. Determine the critical factors, competencies, skills, or bits of knowledge you need to assess. 2. Draw the radar and identify the various spikes by characteristic you are assessing. 3. Determine the scale (numbers often from 0 to 5) and definitions of what each number means, and mark both on the chart. 4. Duplicate the chart, one per person. 5. Ask the group to self-report and mark on the Radar Chart with a date or symbol indicating first measurement. 6. Determine the next measurement point and repeat step 5.

Figure 1. K-Mean Cluster Algorithm (Teknomo, 2006) Two basic concepts used for clustering are dissimilarity and cluster center. As noted before, clustering of data is done by evaluating nearness of data. This means that objects are placed in a topological space, and the nearness is measured by using dissimilarity between two objects to be clustered. A dissimilarity between an arbitrary pair x, x '  X , is denoted by D( x, x ') which takes a real value. This quantity is symmetric with respect to the two arguments:

D( x, x ')  D( x ', x), x, x '  X

Since a dissimilarity measure quantifies nearness between two objects, a small value of D( x, x ') means x and

x ' are near, while a large value of D( x, x ') means x and are distant. In particular, we assume x is nearest to x itself:
D( x, x)  min D( x, x ') x 'X

x'

In relation to a dissimilarity measure, we also have a concept of metric, which is standard in many mathematical literatures. Notice that a metric m(x, y) defined on a space S satisfies the following three axioms: (i) (ii)

m  x, y   0 and m  x, y   0  x  y ; m  x, y   m  y, x  ;

7. Analyze data and compare results, if appropriate. You can choose to do a cumulative average of the group by adding and averaging the entire group’s scores.

3. System Development Method

information flow can be a flowchart and graphics as well as narrative explanations of processes and data. While it functional requirements (related to the software features to be made) according to demand and priority system the parties involved (O'Brien, 2002). In this intelligent business system, the methodology used was the K-Means clustering and the radar chart. The methods intended to provide recommendations to customers. 3.3. System Design McLeod (2002) stated that the system design is the determination of processes and data required by the new system. At the stage of system design method used is UML (Unified Modeling Language). At this stage made a variety of diagrams needed in system development. Diagrams that are needed include: a. Use case diagram (case diagram) This diagram is used to describe a functional system from the user point of view. This diagram emphasizes what the system. This diagram also illustrates the system interaction with the actors (actor) outside the system. b. Activity diagram (activity diagram) Activity diagram is used to describe the work flow activity in the system or in other words, is how the system was doing certain functionality. c. State chart diagrams (state diagrams) Describe how an object changes the status of the trigger events. d. Class diagrams (class diagrams) Class diagram is the main diagram in object-oriented modeling. Class diagrams are used to show the structure of the system status. Class is a collection of objects that have attributes and behavior (operations) that are similar. e. Physical Data Model (physical data model) According Halimsetiawan (2009), physical data model is the number of tables to describe data and relationships between these data. Each table has a number of columns where each column has a unique name. Physical data model is a computer specialist consumption which includes details of data storage in computers. In this concept of data represented in the form of record formats, record ordering, and access paths. 3.4. Application System Stage of implementation of the system include the provision of hardware, software, software development, testing programs and procedures, documentation and the selection conversion alternatives (O'Brien, 2002). Hardware used to operate the E-Cocotrade system is AMD Athlon (tm) Neo L335 X2 Dual Core Processor 1.60 GHz, 2 GB RAM, and operating system Windows 7 Home Premium 32-bit, 500GB external hard drive, CD / CDRW, internet modems, mouse, and printer. The software implemented E-Cocotrade system at this stage of system design using Microsoft Visio 2007 (Microsoft, 2007) and Sybase Power Designer 15.3 (Sybase, 2010) while in the stage of making a program using the software package Adobe Dreamweaver CS4 (Adobe, 2008) and Adobe Photoshop CS3 (Adobe, 2007) to design the interface. MySQL (Oracle, 2009) is used as dynamic database management system. The software required is an internet browser Mozilla Firefox or the equivalent. Testing program use a local server (localhost/xampp). The testing program includes testing website performance, testing of software errors and hardware testing. 3.5. Evaluation System a. Verification

Start Source: -Coconut Company Dekindo and APCC -Books, Journal, Internet

Collecting Data

Information Requirement System Analyze Organizational Analyze Functional Needs Development System Design Output: Use Case Diagram Activity Diagram Statechart Diagram Class Diagram Physical Data Model (PDM) System Implementation Dreamweaver CS4 Adobe Photoshop CS3 MySQL Ms. Visio 2007 Power Designer 15.3 Output: E-Cocotrade

Unified Modeling Language

Website Testing Software Testing

Hardware Testing

System Evaluation : Start/End : Process Appropriate : Decision

-

+ : Data End

Figure 2. Flowchart of E-Cocotrade System The order of the flowchart research in general are: search and secondary data collection, system analysis, system design, system implementation, and evaluation systems. Chronology of the details of governance can be seen in Figure 2 with the following details: 3.1. Searching and Secondary Data Collection At this stage, literature searches related to basic principles of e-business, intelligent e-business, Recommender systems, K-Means clustering method, radar charts, and ways of making e-business, as well as studies on coconut products and derivatives. Source data library derived from books, journals, articles, thesis, and the Internet. Sources of data for coconut and its derivative products will be based on data APCC (Asean Pacific Coconut Community) and Dekindo (Coconut Board of Indonesia). The content of the data used is the name of the company of coconut products, address, email, phone number, contact person, and the company’s product. Data coconut products also obtained from the Internet with content product names, specifications, prices and photos of products sold. 3.2. Systems analysis Stages of system analysis include organizational analysis which obtained the parties involved and their activities. The next step is to analyze information needs and develop functional requirements. In the analysis of information learn the information requirement by users is needed to perform information gathering. Documentation of

According to Hoover and Perry (1989) in Didi (2010), verification is the process of checking whether the operational logic model (computer program) according to the logic flow chart. A simple sentence is whether there are errors in the program. Verification is done to check the translation of conceptual simulation model (flowcharts and assumptions) into a true programming language (Law and Kelton, 1991 in Didi, 2010). The verification process is carried out during manufacture and after completion. Verification stage is a stage that serves to determine whether the program / model that have been created successfully produce the desired output. Stages do is testing and tracking system errors (testing and debugging). b. Validation Validation is the process of determining whether the model, as a conceptualization or abstraction is meaningful and accurate representation of the real system (Hoover and Perry, 1989 in Didi, 2010). Validation is determining whether the conceptual simulation model (as opposed to a computer program) is an accurate representation of the real system being modeled (Law and Kelton, 1991 in Didi, 2010).

IV. DISCUSSION 1. System description Intelligent e-business system for export oriented coconut industry created called E-Cocotrade. This system is Internetbased intelligent systems designed using UML (Unified Modeling Language) to provide product information and giving coconut products advice as a business service for customers. Actors involved in E-Cocotrade are the distributor (PT. Cocotrade Indonesia), prospects and customers. Prospects are individual users or companies that have not been registered as a member of PT. Cocotrade Indonesia. Customers are individuals or companies that have been registered as a member of PT. Cocotrade Indonesia. 2. Result of Clustering Several data of coconut products collected from APCC (Asian Pacific Coconut Community) and coconut industries in Jakarta-Bogor-Tangerang-Bekasi. The products are displayed in product menu as the product available. In this E-business, recommendation system is an important application. In these applications, sets of customers with similar behavior need to be identifed so that we can predict customers' interest and make proper recommendations. Let us consider the following example. Several cutomers buy products of a particular type (VCO, Nata de Coco, etc.) with a different of amount and brand. The customers’ tansaction are stored in the transaction database. Every type of products have a several brands differentiated according to the product id. The customers also can rate the products by giving star value from 1 star up to 5 stars for every criteria. There are 5 rating criteria: design, price, brand, product reliability, and product availability (stock on hand). The products’s rating are stored in the rating database. The clustering method will calculate the distance between the product’s rating and product’s amount. In the future, if the next customer buy or rate the product, the database and the cluster will be changed. According to (Jianxin Jiao 2006), the distance between any two functional requirement instances indicates the dissimilarity between them. In order to take advantage of commonality in product family design, existing instances of

functional requirements should be analyzed and clustered according to their similarity (Tseng 1998). In Table 1, fifteen products are clustered in to three groups which have three centroids. After calculated the dissimilarity with Euclidean metric then the product grouped based on the minimum distance of the products. The Euclidean calculation will end until the last iteration matrix is same as formerly iteration. The amount attribute generate from database transaction. While the design, price, brand, reliable, and the stock is the attributes from the database rating. Coconut products will be grouped into three different groups. C1 is the first calculation of the centroid obtained from calculations using the Euclidean formula. Similarly, C2 and C3 are respectively the centroid calculation of the second and third centroid obtained from the Euclidean calculation formula as well. Then the minimum value at each centroid is presented in matrix form. Value of the matrix 1 will be given to the minimum value and a matrix 0 for other values. Calculation process will continue and stop if the last iteration matrix value equals the value of the matrix in the previous iteration. The result is three clusters noted as high recommend, medium recommend, and low recommend. High recommended products are the product with id code 4, 5, 8, and 15. Cluster medium recommended obtained by the product with id code 2, 12. Low recommended cluster obtained by the product with id code 1, 3, 6, 7, 9, 10, 11, 13, and 14. At E-Cocotrade, data will frequently change because the input of transaction data and the input of rating data. Clustering process is based on data that is frequently changing according to consumer behavior called web mining (Table 1). Table1. Result of K-Means Clustering
ID 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Amount 3 10 5 25 41 0 3 34 1 5 6 10 2 5 43 Design 5 4 3 2 4 5 2 3 2 3 3 5 3 3 5 Price 5 4 4 5 4 3 5 3 5 4 3 5 3 3 5 Brand 4 3 2 3 5 3 5 3 5 3 5 5 5 5 3 Reliable 2 3 1 2 5 2 3 3 3 3 2 3 5 4 5 Stock 5 3 4 3 2 3 2 4 2 3 1 4 3 2 4 C1 3,17 6,81 3,46 21,8 37,8 4,12 2,07 30,7 3,1 2,04 3,54 7,15 2,9 2,55 39,8 C2 32,9 25,8 30,9 11 5,77 35,9 32,9 2,5 34,9 30,8 30 25,9 33,8 30,8 7,6 C3 7,26 1,32 5,98 15,3 31,1 10,2 7,66 24,1 9,53 5,36 5,36 1,32 8,59 5,81 33,1 1 0 1 0 0 1 1 0 1 1 1 0 1 1 0 Matriks 0 0 0 1 1 0 0 1 0 0 0 0 0 0 1

0

1

0

0

0

0

0

0

0

0

0

1

0

0

0

Low Medium High

3.Result of Radar Chart In radar chart, a point close to the center on any axis indicates a low value. A point near the edge is a high value. In Figure 3, the alternative is activated carbon as a product of coconut industry and the criterias are: amount, design, price, brand, product reliability and stock on hand. From the six criteria, the customer is rating the product that his/her had purchased and the rating will be present in the product recommendation database. The product recommendation database connect to the product database, so each product purchased is having spider chart as the recommendation system to help the formerly customer making decision. In Figure 3, the design criteria is less than the price and the brand equals with the stock on hand and amount while the product reliability better than the brand. The following statistics also calculated: 1. Mean as the average of all the values in the series. 2. Maximum as the maximum value in the series. 3. Minimum as the minimum value in the series. 4. Sample size as the number of values in the series. 5. Range as the maximum value minus the minimum value. 6. Standard deviation indicates how widely data is spread around the mean

store, discuss on forum, log in, order, edit account details, view account details, recommendation, maintain inventory, generate report, maintain user, report invoice, pay fee, validation, and give product data (Figure 4).
Sign Up

View Store

Customer Applicant

Comment

Discuss on Forum

Pay Fee Credit Card System Validation

Log In Order Customer Edit Account Details Give Product Data View Account Details

Distributor

Recommendation Maintain Inventory

Generate Report Admin Maintain User

Report Invoice

Figure 4. Fragment of Use Case Diagram for Entire System 4.2. Activity Diagram Activity diagrams are a technique to describe procedural logic, business process, and work flow. In many ways, they play a role similar to flowcharts, but the principal difference between them and flowchart notation is that they support parallel behavior (Fowler 2004). An activity diagram is a special form of state machine intended to model computations and workflows (Fowler 2004). The states of the activity diagram represent the states of executing the computations, not the states of an ordinary object. Normally, an activity diagram assumes that computations are preceded without external event based interruptions. An activity diagram contains activity states. An activity state represents the execution of a statement in a procedure of the performance of an activity in a workflow. Instead of waiting for an event, as in a normal wait state, an activity state waits for the completion of its computation. When the activity completes, then execution proceeds to the next activity state within the graph. A completion transition in an activity diagram fires when the preceding activity is completed. Activity states usually do not have transitions with explicit events, but they may be aborted by transitions on enclosing states (J. Rumbaugh 1999). An activity diagram may also contain action states, which are similar to activity states. Except that they are atomic and do not permit transitions while being active. Action states should usually be used for short bookkeeping operations. Figure 5 explains the activity diagram of recommendation system. The visible actors in the swimlanes are customer, admin, and distributor. The flow of recommendation starts from customers who commit order and end with the use of recommendation decision. The admin use two methods to give recommendation, they are: spider chart method to calculate rating based on criteria and K-

Figure 3. Fragment of Radar Chart

4. UML (Unified Modeling Language) System design method used in this system is the UML. Diagram required in this system includes: 1. Use Case Diagram, 2. Activity Diagram, 3. State Chart Diagram, and Class Diagram 4.1. Use Case Diagram According to the UML, a use case represents a variety of scenarios that can result from the interaction between a system and its environment. Its description has to cover all possible scenarios. Since it is hard and often impossible to name and describe each of them in isolation, a use case is often described in an incremental fashion: One scenario is described completely and explicitly, and all other scenarios are described implicitly in terms of their differences to the first one. UML use case diagrams describe, in the form of action and reaction, the system’s behavior from the user’s point of view. They allow defining the system’s limits and the relationships between the system and the environment (Gaertner 2000). The use case diagrams represent use cases, actors and the relationships between the use cases and the actors. This system use 5 actors and 16 use cases. The actors are: customer applicant, customer, admin, credit card system, and distributor. The use cases are: sign up, view

Means clustering method to grouped the product calculated on Euclidean Metric based on product’s dissimilarity.
Customer Admin Distributor

Commit Order

Buy Purchase Comment on Product Record Order

Record Data Transaction 1

[yes]

calculate rating

Transform Data Transation

[no]

clustering data transaction

generate rating

associate data

Get Reccomendation

Give Recommendation

rating. On the contrary, the customer will return to the commit order activity. (5) Forking and joining: A fork may have one incoming transition and two or more out-going transitions. A join represents the synchronization of two or more concurrent flows of control. In Figure 5, there are one fork and one join. One fork after the commit order activity and one join after the generate rating and associate data activity. (6) Object flow: An object flow state represents an object that is the input or output of an activity (Fowler 2004). There are 21 flows in Figure 5. 7) Swimlanes: There are 3 swimlanes shown by 3 actors regions which grouping together all the activities separated by lines in the diagram (Figure 5). 4.3. State chart Diagram State chart diagrams in UML are used to describe the dynamic behavior of a class, subsystem or system. The key elements in a UML State chart diagram are states, transitions, events and actions. States and transitions define all possible states and state changes that an object can achieve during its lifetime. State changes occur as reactions to events received from the object’s interfaces. Actions correspond to internal or external method calls.
Rating [fill rating] entry/ rating ... [next] Calculated [error] Cancel entry/ record cancellation ... [success] [cancel]

Use Recommendation

[yes]

2

[no]

Figure 5. Fragment of Activity Diagram in Recommendation System An activity diagram may contain branches, as well as forking of control into concurrent threads. Concurrent threads represent activities that can be performed concurrently by different objects or persons in an organization. Frequently concurrency arises from aggregation, in which each object has its own concurrent thread. Concurrent activities can be performed simultaneously or in any order. An activity diagram is like traditional flow chart except it permits concurrent control in addition to sequential control-a big difference (J. Rumbaugh 1999). The activity diagram contains seven types of state: 1) Action state, 2) Initial/Final action state, 3) Transition, 4) Branching, 5) Fork and join, 6) Object flow, and 7) Swimlanes. (1) Action states: In the flow of control, action stated is modeled by an activity diagram. It might evaluate some expression that sets the value of an attribute or returns some value. In Figure 5, there are 13 activities: 1) Commit order, 2) Comment on Product, 3) Record Order, 4) Buy Purchase, 5)Record Data transaction, 6) Transform Data Transaction, 7) Clustering Data Transaction, 8) Associate Data, 9) Calculate Rating, 10) Generate Rating, 11) Give Recommendation, 12) Get Recommendation, and 13) Use Recommendation. These executable, atomic computations are called action states because they are states of the system. Each represents the execution of an action. (2) Initial/Final action state: The initial action state can express the first action state in activity diagram and the start state is represented by a solid cycle. The final action state can express final action state in activity diagram and the final state is represented by a concentric cycle. The first and the final state in this diagram (Figure 5) are represented in customer’s regions. (3)&(4) Transitions & Branching: According to G. Booch (1999), transitions is used to show the path from one action or activity state to the next action activity state. A branch, which specifies alternate paths taken based on some Boolean expression. It represents a branch as a diamond. In Figure 5, there are 2 branches (diamond number 1 & 2). The diamond number 1 shows boolean expression of activity (comment on product) which conditioned two outgoing transition. If the customer comment on the produt, the admin will calculate the

Request Completed entry/ recommendation ... [complete]

[uncomplete]

Radar Chart do/ recommendation ... [continue] [success] Recommendation

Figure 6. Fragment of State Chart Diagram in Product Rating For example, Figure 6 is the State chart of recommendation. It comprises six states, namely Rating, Calculated, Cancel, Request Completed, Radar Chart, and Recommendation respectively, as well as a start and an end state, denoted by and . There are four actions in recommendation. The transitions from source state to target state are labeled with “event [guard condition]/action”, which means that, when source state detects the event and guard condition is also satisfied, a transition will occur by executing action. They are entry/rating, entry/record cancellation, entry/ recommendation, and do/recommendation. 4.4.Class Diagram Classes are depicted as boxes with three sections, the top one indicates the name of the class, the middle one lists the attributes of the class, and the third one lists the methods. Figure 7 explains the recommendation class which connected to the customer class, product recommendation class, product class, and admin class. The attributes of recommendation class are the rate of design, quality, price, brand, producer company, product reliability, stock/continuity, and people opinion. This class will generate into PDM then databases. The databases of the recommendation will insert by the previous customer and

select into the product recommendation class which give the recommendation to the next customer by showing the product’s radar chart.

given will affect the radar chart and the clustering of each product. Value of rating on the rating database will continue to change every replenished by the next customer.

Recomendation Id Product Product Name Produsen Company Spec Selling Price Product Photo Rate by Design Rate by Quality Rate by Price Rate by Brand Rate by Produsen Company Rate by Product Reliability Rate by Stock/Continuity Rate by People Opinion Comment : : : : : : : : : : : : : : : int String String String int byte int int int int int int int int String

0..* Line Recommendation

+ calcRate () : int ... 0..* Line Product Rec 0..1 0..1

Product Recommendation 0..* Line Order Product Product Id Product Name Produsen Company Spec Stock Purchasing Price Selling Price Product Photo Rating Product : : : : : : : : : int String String String int int int byte int

Figure 9. “Give Rate” Link Interface In addition there are also links “radar chart” on each product. This link is used to display the customer's radar chart of rating’s result which show value of the criteria: design, price, brand, product reliability, product availability (stock on hand) and the value of products sold (amount) taken from the transaction database. Amount’s value here means the amount (in percent) of products sold. If the total amount of the product sold above or equal to 50%, it will be shown 5 stars. If between 40% -50%, it will be shown 4 stars. If between 30% -40%, it will be shoen 3 stars. If between 20% -30%, it will be shown 2 stars. If between 10% -20%, it will be shown a star. If below 10%, then there will be no stars shown. Display radar chart and the explanation can be seen in Figure 10 below.

Contact Contact Id Name Address Phone E-mail : : : : : int String String int String

1..1 Line Contact

+ calcPrice () : int + getRate () : int ... 0..1

0..* Data Customer

0..1 Customer 0..* Line Customer 0..* Line Product Product Product Id Product Name Spec Stock Purchasing Price Selling Price Product Photo Rating Product : : : : : : : : int String String int int int byte int

-

Customer Id First Name Last Name Email Address Password Date of Birth Phone Number Fax Number Gender Photo Company Name Address Zip Code CIty Province Country

: : : : : : : : : : : : : : : :

int String String String String int int int String int int int int String String String

0..1 0..1

0..* Line Product

+ calcPrice () : int + getRate () : int ...

Figure 7. Fragment of Class Diagram in Recommendation System 5. E-Cocotrade Program Recommended menu display the products from the clustering database which have calculate the centroid with euclidean metric. (Figure 8). In this menu users can select products based on product groups: high recommended, recommended medium, and low recommended. In the programming language the database generate the database transaction adn database rating. Then iterated the data until the data matrix equal to the previous matrix.

+ authorized () : int + get recommendation () : int ...

Figure 10. “Radar Chart” Link Interface Customer Services menu has a discussion forum. Customer Services menu became the hallmark that distinguishes between e-commerce with e-business. In ebusiness there is a form of CRM (Customer Relationship Management). Customer Services menu is one of the CRM application, customers can conduct a discussion here with other customers. The topic has been visible in the menu, customers can also add a new topic which then can later be returned by another customer. Customer Service menu can be seen in Figure 11.

Figure 8. Recommended Menu Interface Display recommendations will change according to changes in the database and database transaction rating. More customers who buy products and more high ratings are given the product will be good product recommendations for customers (high recommended). On the products menu there is a link to give rate the products located on each product line. Customers used this link to rate products with the following criteria: design, price, brand, product reliability, and product availability (stock on hand). Figure 9 is a display of the link “give rate”. Rating Figure 11. Display Menu Customer Service

After discussing the display or the frontend user interface, we will be discussed the display interface administrator or commonly called the backend. The structure of the administrator interface consists of the initial page the index administrator. On the home page administrator have to input the username and password to proceed to the administrator page. On that page there are seven menu selections are: Sales Graph, Orders, Manage, Recommendation, Comment, Users, Sales and Profit. Sales Graph on the menu the administrator can view the sales report every month that are presented in bar graphs. Orders on the menu there are two sub-menu options, Transaction menu and Shipment menu. Administrators can view reports of sales transactions in the sub Transaction menu and data delivery on the Shipment menu. On the Manage menu there are three sub menus: Product, Industry, and Delivery Cost. Administrators can update the product data on the sub menu Product, updating data on industrial enterprises Industry submenu, and update the shipping costs Delivery Cost on the menu. On the menu Recommendation terapat three sub menus: Ratings, Radar Chart, and Clustering. Administrators can check product rating Rating on a sub menu, check the radar chart in Radar Chart sub-menu, and check on the sub menu clustering Clustering. Comment on the menu the administrator can check user comments and remove the data. On the Users menu, there are two sub menus: Customer and Admin. Administrators can update the user accounts on the sub menu Customer and renew an administrator account on the Admin sub-menu. Profit on Sales menu the administrator can display product sales profits. In the administrator view (Figure 12), provided a box to enter a user name and password.

Sales Graph on the menu, the administrator can see in the form of monthly charts in a matter of years. Display Sales Graph as shown in Figure 14.

Figure 14. Display Sales Graph Display Sales Graph above using the graph module is inserted in the htdocs folder and change the variables according to the variables that exist in the database transaction. V. CONCLUSION The study produced the design in the form of simulated E-Cocotrade program package, a package of simulation programs used by PT. Cocotrade Indonesia (the company made) for oil palm agro-industry. Application of intelligent e-business on this system is the recommendation system using the radar chart and clustering. These recommendations are grouped according to customer preferences in the rate of criteria: design, price, brand, product reliability and stock on hand. Clustering helps customers to make decisions by providing the best product recommendations from all products. There are three clusters in this program, namely: high recommended, recommended medium, and low recommended. Furthermore, the radar chart gives the comparison criterion function on each product to display the rating charts and explanations. The difference of e-business with e-commerce in this study is the existence of customer service. On the customer service there is a forum for discussion between the customer and providing comments to the administrator. Discussion forum serves as a means of communication between the customer E-Cocotrade while providing commentary serves as a means of communication between the customer taking into administrators. Facilities owned by PT. Cocotrade Indonesia are generally divided into two interfaces, the customer display interface (frontend) and display interface the administrator (backend) to enhance system security. The information available to the administrator is setting product data, user data, cost data delivery, order status, check transactions, as well as the calculation of profit per month per product item, the value ratings of products and also information sales charts per year. Advice that is used for the implementation and development of e-business further research are: 1. Published (uploaded) to the Internet network with www.cocotrade.net name. 2. Cooperate with banks and PayPal Inc. to facilitate the transaction process. 3. The addition of e-business applications such as ERP (Enterprise Resource Planning) 4. Develop an online store system to auction sites (auction website).

Figure 12. Initial View Administrator On the administrator menu, there are seven menus: Sales Graph, Orders, Manage, Users, Recommendation, Comment, and Sales Profit. Administrator menu display can be seen in Figure 13 below.

Figure 13. Administrator menu

VI. REFERENCE Adobe. 2007. Adobe Photoshop CS3. Adobe. 2008. Adobe Dreamweaver CS4. Ahn, K.-j. K. a. H. ( 2006) A recommender system using GA K-means clustering in an online shopping market. Didi. 2010. Verification and Validation Model Systems. http://didi.staff.gunadarma.ac.id/Downloads /files/5040/VERIFIKASI+DAN+VALIDASI+MOD EL+SIMULASI.pdf. [5 Jun 2011] Fowler, M. (2004). UML distilled: a brief guide to the standard object modeling language. Boston, , MA: Addison-Wesley Publishing. Friendly, M. (1991, Apr, 1991). "Statistical Graphics for Multivariate Data." from http://en.wikipedia.org/wiki/Radar_chart#cite_ref-3. G. Booch, J. R., and I. Jacobson (1999). The Unified Modeling Language Reference Manual. Boston, MA: Addison-Wesley Publishing. Gaertner, P. A. M. a. N. (2000). Modélisation Objet avec UML. Paris. Halimsetiawan JH. 2009. Conceptual Data Model (CDM) and Physical Data Model (PDM). http://tutorialpemrograman.wordpress.com/2009/08/ 08/konsep-conceptual-data-model-cdm-danphysical-data-model-pdm/ J. Rumbaugh, I. J., and G. Booch (1999). The Unified Modeling Language User Guide. Boston, MA:Addison-Wesley Publishing. Jianxin Jiao, Y. Z., and Martin Helander (2006). "Analytical Customer Requirement Analysis Based on Data Mining." Idea Group Inc. McLeod R. 2002. Management Information Systems. New Jersey: Prentice Hall. Microsoft. 2007. Microsoft Office. NIST/SEMATECH (2003, 23 March 2011). "Star Plot in: eHandbook of Statistical Methods." from http://en.wikipedia.org/wiki/Radar_chart. O’Brien (2002). Management Information Systems. New York, Mc Graw Hill. Oracle. 2009. MySQL. Oracle Rogers, C. B. ( 1995). The Spider Chart: A Unique Tool for Performance Appraisal, ASQC. Sadaaki Miyamoto, H. I., and Katsuhiro Honda (2008). Algorithms for Fuzzy Clustering. Methods in cMeans Clustering with Applications. Verlag Berlin Heidelberg, Springer. 229.

Sybase. 2010. Power Designer Studio Enterprise Standalone Local. 15.3.0.3248 ed. Teknomo, Kardi. 2006. Numerical example of K-Means clustering. http://people.revoledu.com /kardi/tutorial/kMean/NumericalExample.htm. [14 Mei 2011] Tseng, M. M., & Jiao, J (1998). "Computer-aided Requirement Management for Product Definition: A Methodology and Implementation." Cocurrent Engineering: Research and Application 6: 145-160. Wanurmarahayu, C. (2011). Design of Internet Business System (E-Business) for Leather Agroindustry. Department of Agroindustrial Technology. Bogor, Bogor Agricultural University. Zeng, Z. (2009) An Intelligent E-Commerce Recommender System Based on Web Mining. 4, 10