Wednesday, 15 June 2016

Recommendised E-Trade Through Web Data Mining
Dr. Sandeep Gupta
Department of Computer Science and Engineering, JIMS Engineering Management Technical Campus, Gr.noida (affiliated to GGS IP University, New Delhi, India)
Sandeep.gupta@jagannath.org
   I.     INTRODUCTION

The aim of individualizing information services is to provide users with individually tailored news and information. It is to optimally support them retrieving information within their scope of interest in the desired coverage, access time, and media. Personalization is also about making online information easier to access, more efficient in use, and providing an individual, personal online experience for the users. Thereby personalization can be considered twofold: as a feature that supports the information supply of a particular information provider, or as an autonomous concept which is to grab and filter information and contents from several information sources and individually present it to the users.
In either case the major goal is to help the users getting the information they want in an efficient and pleasing way. Yet the hurdle with achieving this goal is that the information system must first know the users’ interests, needs, and preferences. Many of the existing individualizing services put the strain of gathering this information on the users. Users are confronted with incomprehensive registration forms, prompting the users to explain their life story in data. Next they have to bother with seemingly endless individualization and customization features, having to choose from dozens of check boxes or enter cryptic search and keyword strings.
The result is that many users either give up before having finished the individualization process or they end up being frustrated because the effective results often they stay behind the expectations of the users. Such systems seem to ignore that personalization is not the actual goal task of the users but rather just getting the information they are interested in.

II.       RECOMMENDISED E-TRADE

The recommendised information service has its roots in the personalized navigation system raised by Robert Armstrong and others of Carnegie Mellon University at the National Conference on Artificial Intelligence (AAAI) in 1995.

Recommendised information services are now an emerging research topic in global information services. Conventional information services are giving gradually an approach to individualize information services that, according to the interest, status and customized needs of a customer.

Recommendised assistance in e-trade mean that in the mode of e-trade a entrepreneur obtains his customer’s data and access information by depending on his customers’ access to entrepreneur website, and examines and exercises such information by using web data mining tools to guide his enterprise decision so that based on the requirements of his customers the entrepreneur could embark on e-trade activities, offer recommendised information services, improve the awareness , satisfaction and loyalty of the users and gain win-win for the entrepreneur and his customers (as shown in figure1)


Figure1. Methodology for Recommendisation

Recommendised services in e- trade, in nature, are web services centered on individualize requirements. In Figure 2, where “User Analysis Module” functions to learn the user’s features, create user access model and offer recommendised services to the user by using technology processing and web resources.
Text Box: Resource
Scheduling



     Information
  Feedback

E-Trade Recommendised Service Platform
 



User Analysis   Module          
 
             User Access          
           
 
    User
           Individualized Services


Figure 2. User Analysis Module - Meaning of Recommendised Services in E-trade

 
III.       Web Data Mining

As a key technology to provide recommendised e- trade and help to collect individualize information, web data mining can be used to analyze and examine user data, create access model, requirement model and interest model that accord with user features, making recommendised e-trade possible.
Oren EtZioni put forward the concept of web data mining in his thesis in 1996 for the first time: applying data mining methods to web helps to discover potential and useful models and information.
According to data mining nature, web data mining is classified as web content mining, web structure mining and web usage mining. For more information, see Table 1. Web usage mining means that by mining the log files and data at the corresponding site, we discover the nature of visitors and users having access to this site. Data mining methods include path analysis, association rules, classification rule, sequential patterns, statistical analysis, dependent relationships modeling and cluster analysis




Classification
Secondary Classification
Web Content Mining
Text Mining
Multimedia Planning
Web Structure Mining
Organizational Structure Mining
Page Structure Mining
Web Usage Mining
User Record Data Mining
Customization Mode Mining
Table I. Classification of Web Data Mining

Data on web is unstructured, semi-structured and dynamic, so web data mining has to go through the corresponding processing flow that is composed of data positioning, data preprocessing, pattern recognition and pattern analysis. In this process, we should first determine the source of data, including web document, e- mail, website log data and transaction data. Next, we should preprocess data, i.e., delete some redundant information and unify information recognition, session recognition and transaction recognition, and then carry out pattern recognition of the preprocessed data, i.e., use the data mining tools to extract useful, potential and understandable information. Finally, through pattern analysis, we can convert the filtered data into useful rules and patterns to guide the practical e-trade activities.


No comments:

Post a Comment