Finding Association Rule In Ecommerce Using Fuzzy Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Today Electronic commerce is very popular for marketing in every business environment. It is convenient way for users to purchase or sale items using internet. When a user click on the particular link in a website, this information is easily stored in the weblog database file and when a user purchase or sale items from the website, this transaction detail are stored in the user transaction database. These two different databases can be used to generate strong association rule to develop effective marketing strategy for business growth. In this paper, we design architecture for association rule generation using fuzzy logic, which uses two different databases like user transaction and web log to produced strong fuzzy association rule. We also compare this approach with the basic association rule mining Apriori algorithm, in which we found efficient results.

Index Terms- Fuzzy Membership function, E-commerce, Association Rule, Web Mining.


Data mining can be defined as a process for extracting useful knowledge from the databases. It also uses for finding useful and unknown knowledge information from the large databases [2].Data mining uses different methods and techniques for knowledge extraction. Method or techniques selection depends on the category of data or knowledge needs. Data mining performs different kinds of tasks like association Classification, clustering, users trend analysis of summarization [1] [3].Data mining is very Useful at present time for E-commerce. it provides techniques to market researchers for developing new ideas and marketing strategies for their business. Association rule designing from the transaction database and relationship preparation between attributes of large data is an important task of data mining.

Association rules are used to find unknown relationship and one can forecast and take decision on the bases of these rules. So that effective association rules development is highly preferred area in data mining.

In past, Agrawal and Srikant proposed Apriori association rule mining algorithm [6], which is used to find meaningful itemsets and prepare association rule within large datasets. but this algorithm generates large number of candidate itemsets and it works level by level so it require repeatedly database scanning for each candidate itemset. So that its performance is affected, it is shown in fig. 1[5].

In Literature Review many approaches like DHP algorithm [9], DIC [8], the sampling algorithm [10] and cluster based association rule approach have been proposed to improve the efficiency of data mining process. Agrawal et al. also proposed a method for mining association rule from data sets using quantitative and categorical attributes [7].

Itemsets 2

Itemsets 2

Itemsets 1


Scan 1 in Database D

Scan 2 in Database D







Itemsets n

Scan n in Database D

Scan 3 in Database D

Database Scan in Apriori algorithm

This method first determines the number of partitions for each attribute and prepares maps for every attributes into integer forms. Now these days fuzzy set theory are used frequently for designing intelligent systems because it is mostly similar to the human reasoning [11]. Fuzzy set theory also used by the different researchers. Hong et al. prepare a fuzzy mining algorithm for quantitative data management [15]. This algorithm does not require hierarchical relationship between items. But in real world applications some hierarchical relationships between items are mostly useful. previous research work only focuses on association rule mining and finding items relationships from transaction databases. But few researches also includes some other issues like Tsai et al. prepared graph based algorithm to discover association rule between large items from transaction and attribute values as well from customer database [16].

Today data mining techniques are mostly important in Ecommerce for knowledge discovery from web documents and web services. Web mining provides facilities to automatically extracting information from the web documents. Here data mining techniques are also useful in evaluation and analysis. For example if we extracts web document from world wide web(www) [17] then almost 90% data is useless and not provide relevant data, so that we require data mining techniques for making data useful for us.

Web mining generally addresses unstructured or semi structured data like web and log files, html tags, hyperlinks, audio, video, text or other type of data. This type of unstructured data presents incomplete information. So that fuzzy set theory plays important role to mine useful information from such type of data [14]. In this paper we discuss an important issue of association rule mining. That issue is the discovery of association rule from the web usage data and large itemsets identification from the transaction database. The goal of this paper is to develop a framework for finding association rule in E-Commerce environment. This paper is organized as follows. First, we define the concept of association rule, fuzzy logic, membership function and web mining in section 2. The proposed architecture for association rule generation is shown in section 3 and section 4 shows an illustrative example for this architecture. Conclusions and future works are discussed in section 5.


First association rule method was proposed by the Agrawal and Srikant for discovering large itemsets and constructing association rules named Apriori Algorithm [6]. It is a data mining method and works on support and confidence. In this algorithm repeatedly database scanning requires and generates candidate itemsets. For example, at the (n-1) Th iteration, itemsets contain n-1 items, and then n-1 itemsets are generates. In next iteration the candidate itemset contains n items are generated by joining (n-1) itemsets. When all itemsets are generated then it is require to compute their confidence for constructing association rules.









Grouping Structure Example

Several research works have been done after Apriori algorithm [6] to reduce the number of database scan for itemset generation. Park et al. proposed an effective algorithm known as direct hashing and pruning for initial candidate set generation [9]. Tsay et al. have used cluster based table to reduce the number of database scan and requires less contrast. Most of the literature studies of data mining describe that single concept level hierarchical relationship are used to extract association rules.

In this example the Communication falls into two classes computer and mobile. And mobile can also further classify into Samsung and Nokia. Similarly Electronics can be classified into two classes TV and Radio. In transaction database only terminal items (Samsung, Nokia, TV, and Radio) can appear. Another association rule mining issue is related to find relationship from multiple databases. Tsai et al. design a method for finding association rule using customer and transaction databases. In this work they discovered itemsets from transaction database and then design graph based algorithm, used for finding association rule from both customer and transaction database.

Fuzzy Set

Fuzzy set theory was first proposed by Zadeh in 1965 [13].fuzzy set is something different to the traditional sets, in which every element of the set must either be in or not in a set [27].In a set A, the function assign a value FA(X) to every such that

FA(x) =

So the function maps element of universal set to the set containing 0 and 1. This type of function can assign values to the element within specific ranges. It also describes the membership of the elements within set. Such function called membership function FA(X), which can be defined in fuzzy set as well. This membership function can be represented as FA: X→ [0, 1] where [0, 1] denotes the interval from 0 to 1. And function can be assigning any value between 0 and 1 to the elements [1] [14]. Fig. 3 shows the membership function example.


Min Avg Max


Fuzzy Membership function

Suppose x1 to xn are the elements in fuzzy set A and f1to fn are membership values respectively then A is usually represented as:

The Following equation denotes the scalar cardinality of fuzzy set A on finite universal set X

Union and Intersections are the basic operations of the Fuzzy sets which is proposed by the Zadeh [13].

)Union of the fuzzy set A and B

Intersection of the fuzzy set A and B

Fuzzy logic can be applied on the fuzzy sets to extract knowledge which is imprecise. A present time it is mostly used by the researchers. Hong et al. proposed Fuzzy data mining method to find generalized Association rules [15].

Web Mining

Web mining is very similar to the data mining and text Mining; it deals with the web data. It is used to extract useful information from the web. When we apply web mining on the web data, it is important to apply some Pre-processing and refinement task on that data. These Pre-processing tasks are important because data includes text, images, audio, and video, document, link etc. Web mining is basically used for Resource discovery, Information selection, Generalization of data, data analysis and visualization. it can be classify in following ways [4].


-Pattern Discovery

-Pattern Analysis

-Web Page Content

-SearchPage Content

-Result Page Content

-Links Structure

-Internal Structure

-URL Mining

Web Mining

Web Content Mining

Web Structure Mining

Web Usage Mining

Web Mining

Web Content Mining

Web mining is basically used to get information from the web. When we open a page on web, it includes different contents like hyperlinks, text, multimedia and other things. These are called the contents of web.

Web Structure mining

Web structure mining can be defined as the graph, which is the collection of nodes and edges. Here, web pages are used as a node and hyperlinks as a link. So that web structure represents the relationship between different web pages.

Web Usage Mining

It is the discovery of generated pattern by the client server interaction/transaction. These pattern s are automatically generates when user click on a link, user give queries, from client side cookies, from user profiles and server logs etc.


In this paper, we design architecture for fuzzy association rule generation. The architecture is shown in figure 5.

User Transaction Database

Weblog Data



Fuzzy Document set generation algorithm

Fuzzy set Tables

Document Sets

Fuzzy Rule generation algorithm

Information Database

Fuzzy Association Rule

Fuzzy association Rule Generation Architecture

This architecture is divided in following sections.

Identify the Large Document Sets from the User Transaction Databases.

In this step we scan the user transaction database and create fuzzy set tables using predefined membership function. Firstly we scan the user transaction database and then create fuzzy set tables. This large document set generation process is similar to the basic association rule generation Apriori algorithm. for this we use a predefined fuzzy membership function with three regions low, middle and high. We calculate the summation of each fuzzy region in the transaction data. if the length of transaction record is k, the transaction record and the fuzzy region value of items in this transaction will be stored in the table, named fuzzy-sets table (k), 1≤k≤M, where M is the length of the longest transaction record in user transaction database. Now the set of candidate itemsets Cn is generated by

the self-join of Ln−1.If the fuzzy region value of Cn is greater than or equal to the predefined minimum support Value. The candidate itemsets becomes the large itemsets, put Cn in the large itemsets Ln. finally we use the predefined minimum confidence value to discover fuzzy association rules.

Collect the information from the World Wide Web.

When user click on a particular link in a website, his behavior can identify. This click information can be stored in the web server log files. We use following method for finding web usage information for our purpose-

i. We design a fuzzy membership function and apply preprocessing tasks on the web log data.

ii.We group together the items according to click and also store total clicking frequency for each datasets.

iii. Now we compute membership grade of each frequency, using predefined membership function. And this information is stored in the information database.

Fuzzy association rule generation process.

In this step company can plan out some marketing strategies according by designing relationship between large fuzzy document set, fuzzy itemset and fuzzy frequency of itemsets. Now this relationship, transformed into association rule like "IF user click on Pi THEN buy Qj frequently". This method is applied as follows-

i) We scan information database of frequency membership grade and store set of customers as Ui according to a minimum threshold.

ii) We scan fuzzy itemset table to filter the set of customers for each large fuzzy itemset and store as Lj according to a minimum threshold.

iii) A relation matrix R is created with all pairs of Ui and Lj.

iv) If the value of confidence for each pair (Ui, Lj) is greater than or equal to the user predefined minimum confidence, the relationship is transformed into a fuzzy rule in the form of "IF user click on Pi THEN buy Qj frequently".

Experimental Results

Fuzzy Document Set Generation

In this section we generate fuzzy document set by the various transactions. Table 1 shows an example of transactions with items.

Transaction Table



Transacton ID






















We use same membership function for all items which is shown in figure 6.

Fuzzy membership function for document set generation

This membership function have low, middle and high region. And regions fuzzy membership values are produced for every item by predefined membership function. These are shown in figure 7, figure 8 and figure 9 with different input values.

Rule Viewer with low input

Rule Viewer with middle input

Rule Viewer with high input

Web usage information collection

In this section, we prepare an example to show the information gathering from the web. for example if we have items A,B,C, D and E and they are classify in T1,T2,T3 and T4 Classes. It is shown in table 2.

Transaction Table





CustomerId set





















We also prepare a membership function for finding membership grades of the different frequencies of the transaction from the web log data. This function shown in figure 10.

Fuzzy membership function for web usage information

Fuzzy Association Rule Generation

At first, we use fuzzy document set generation process to find the fuzzy association rule, for this we uses predefined minimum support. So that we find rule like "IF user buy item T1 with middle degree THEN user can buy item B with middle degree together". Next , we calculate the fuzzy membership grades for item frequencies like T1.middle=0.48, T2.middle=0.53, T3.middle=0.55 and so we assume the minimum confidence value 0.5 therefore we got the grade values like T2.middle and we find the rule like "IF user browse class T2.middle THEN buy item B with middle degree".similarly we can produced lots of association rule by this proposed approach. We evaluate the efficiency of the proposed approach with a known Apriori based fuzzy association rule algorithm[18], we matlab on pentium IV pc with 1GB of physical memory to perform the experiments.the experimental result shown in figure 11 and figure 12.

Efficiency of proposed approach

Performance comparison


In this paper we design a framework for business users to plan appropriate decision strategy which is useful for the growth of a business. This study based on the E-commerce data which is generated by the user's website log database and their purchasing transactions. By this framework company managers are able to make strong association rules for more selling of their items. This framework is more efficient to the basic association rule apriori algorithm.