Data Mining for E-Commerce - Research Paper Example

Data Mining for E-Commerce Abstract Data mining is a newly emerged paradigm that has gained a huge recognition regarding the knowledge discovery from complex data sets. In addition, the implementation of the data mining techniques in marketing has offered huge advantages in better customer tracking, buying behavior analysis, market segmentation, targeting customers, and most especially in web based advertisement. This paper presents an analysis of data mining in terms of its implementation for e-commerce systems. The basic purpose of this research is to analyze the use of data mining for e-commerce. This paper will also outline the main areas of implementation, techniques, and potential advantages obtained through this technology. Introduction Data mining is a detailed process which allows the extraction of hidden, formerly unidentified, and actually functional knowledge and information from a huge collection of data. The majority of researchers have defined “data mining as the process of getting useful and reliable information and patterns from huge data sets by making use of latest tools and algorithms based on the theories and models borrowed from various other domains such as machine learning, management systems, statistics and database.” The basic purpose of extracting these hidden facts is to facilitate business executives and top management in planning and managing the business strategies and plans for the future. The use of data mining tools and techniques provides a large number of benefits and opportunities for business organizations. For instance, data mining tools and techniques allow the business organizations to carry out a deep examination of the customer and business associated data and information, which facilitate business firms to make critical strategic decisions. Additionally, data mining applications can be accessed through a graphical user interface (GUI) which helps business managers to take deep insight into the collected customer data. In the past few years, there have emerged a large number of powerful data mining algorithms and techniques to help business managers analyze large customer data sets which are the need of the majority of the business firms for the reason that the survival of their business heavily relies on these data and information (Ranjan & Bhatnagar, 2009). Moreover, the data mining offers these decision making capabilities by making use of a wide variety of methods such as classification, clustering, prediction, genetic algorithms, association and neural network. In this scenario, classification refers to the process of determining the significant attributes and features of customers’ data which are on the point of churn as well as it also helps to identify the customers. In the same way, some clustering techniques such as K-mean algorithms are used to develop segments of this collected data. Additionally, these data are divided into segments on the basis of their features and attributes. In this scenario, the data with same properties is placed in the same set. Hence, this information can be used by a business organization to determine the potential customers of the firm. There is another useful data mining technique known as a prediction technique used to plan the business strategies for the future. In the same way, some association rules are applied to help business firms recognize resemblance among the customer data sets. Some of the important application areas of the association rules can comprise attached mailing in direct marketing, market basket analysis, fraud detection, etc. Another most important data mining technique known as the neural network can be utilized for analytical models for attrition, churn management as well as can also be used to compute the customer lifetime value. Furthermore, at the present these data mining tools and techniques are applied in almost every area of the business (Ranjan & Bhatnagar, 2009). In the past few years, the use of internet for carrying out business tasks has increased to a massive extent. In fact, the majority of business organizations has started supporting their business processes through the Internet. In this scenario, e-commerce is the process of carrying out business tasks using the internet (Laudon & Laudon, 1999, p. 25). When more and more data turn out to be accessible for online public sale, the need for effective data mining techniques turns out to be vital for identifying the hidden intelligence or knowledge hidden in the dataset. Additionally, a large number of data mining tools and techniques have emerged in a wide variety of business departments. In addition, a large number of researches and studies have been conducted to study the impact of data mining techniques on customer relationship management (CRM) and other marketing fields. However, only a few researches have been conducted to discuss the implementation of data mining tools and techniques for building the most efficient selling and purchasing strategies for e-commerce sites (Tu, 2008). In addition, one of the most significant aspects of e-commerce sites is to be capable of providing a customized experience for the visitors and customers. At the present, there has been more and more research concern in forthcoming with tools and techniques to develop sensible customers’ usage profiles. In this scenario, the developments of profiles with respect to the use of a service and information available on e-commerce site allows the software developers of the e-commerce site to redecorate e-commerce site or a service in accordance with the actual requirements of the web site visitors or customers and even offer customized browsing experience for each one. In fact, one of the biggest and most successful e-commerce giants Amazon.com follows the mantra “a customized e-commerce site for every customer”. In this scenario, for this usage profiling a well-known data mining technique is used, which is known as web-based data mining. Basically, this technique allows e-commerce businesses to gather information and useful knowledge from the usage patterns of an e-commerce site (or a particular service) with the purpose of afterward evaluating it as well as identifying some interesting usage patterns. In simple words, these particular data mining techniques for discovering the web based usage patterns are usually covered under a separate domain known as web usage mining, and balanced with what is acknowledged as web content mining and web structure mining (Navarro-Arribas & Torra, 2010; Facca & Lanzi, 2005). This paper presents an analysis of data mining in terms of its implementation for e-commerce. This paper discusses how data mining can be used in e-commerce. This paper also presents recommendations that should be kept in mind while implementing data mining tools and techniques for e-commerce. Data Mining and E-commerce In the past few years, a large number of researchers have devoted them for web mining research. The basic purpose of carrying of these researches and studies is to put into practice or enhance customary data mining practices with the purpose of adapting them to the modern ways for data source and analysis target. Additionally, the majority of researchers takes into consideration the development and implementation of the latest and modern web mining tools, algorithms and practices. On the other hand, there is no value of such work if these tools, algorithms and practices cannot be applied in a real environment. In addition, e-commerce and e-services come under the research area of web research and they have been recognized as significant domains for web mining approaches. The research has shown that web mining approaches play a significant role in e-services and e-commerce. In fact, it has been proved to be successful technique for identifying how e-service, e-commerce websites and services are accessed and used, allowing the business organizations to design and put into practice better services for general users as well as regular customers (Ting, 2008; Zhang & Segall, 2010; Lappas, 2008). Web-mining taxonomy Data mining tools and algorithms designed for e-commerce are studied under the domain of web mining, which is the application of data mining tools and algorithms to the Internet and its basic aim is to recognize business intelligence from the web sites (particularly the e-commerce web sites) (Tu, 2008). In addition, web mining is divided into three mining classes in accordance with a variety of data sources: Web-content mining In this data mining technique knowledge from the content of web pages is discovered. In this scenario, the target data are composed of multivariate types of data that is integrated in a web page in the forms of images, text, multimedia, etc. (Zhang & Segall, 2010; Lappas, 2008). In addition, web content mining is more than deciding for applicable documents on the web. Additionally, web content mining involves the process of extracting useful information and knowledge from evaluating a huge set of web documents. In view of the fact that a web site contains a wide variety of multimedia sources in the forms of images, audios, videos and animations along with normal text documents such as HTML pages and links, hence the content indexing of this multimedia content is much more complicated than simple document indexing, creating the need for “multimedia web mining”. In this scenario, multimedia web mining is a new web mining paradigm which is aimed at performing mining operations on multimedia files contained in the web with the purpose of the extraction of useful information and knowledge embedded in multimedia web data (Lappas, 2008; Kosala & Blockeel, 2000; Cooley, Mobasher, & Srivastava, 1997). Web-usage mining In this type of data mining knowledge from user navigation data is discovered while a website is being used. In this scenario, the target data comprises the requests from users which are stored in individual files and kept on the website's servers, which are known as log files (Zhang & Segall, 2010; Lappas, 2008). The basic purpose of web usage mining is to determine interesting patterns of navigational patterns from users visiting an ecommerce site. In addition, these useful patterns can be helpful in providing the answers of the following questions: How efficiently an ecommerce web site provides the desired information? What do users feel about the structure of an ecommerce web site? Does this information can be used to predict the user's next visit? Can this website be improved to meet user needs? Is there need to make changes to web site with the purpose of increasing users’ satisfaction? Can this web site be used to target specific groups of users and customize web content for them? Moreover, the answers to these questions can originate from the detailed analysis of data contained in log files which are kept and maintained in web servers. The research has shown that web usage mining has turned out to be an essential activity with the purpose of providing web managers with significant information on the subject of users and usage behaviors and patterns for increasing the level of the quality of both the service performance and web information. It is an admitted fact that successful ecommerce web sites are those which can be personalized in order to get together their users’ desires both in the production of information as well as in the application of the content to the user's needs. In this scenario, these discovered usage and user preferences patterns can be utilized in targeting particular groups of users in a variety of recommendation systems and in the assessment and rebuilding of the website to get together design effectiveness problems as well as customers pleasure requirements (Lappas, 2008; Kosala & Blockeel, 2000; Cooley, Mobasher, & Srivastava, 1997). Web-structure mining This type of data mining focuses on the connectivity of websites and the knowledge and information discovery from hyperlinks on the web (Zhang & Segall, 2010; Lappas, 2008). Additionally, web structure mining algorithms are applied in modern search engines to establish the position of the application of websites as well as to categorize them on the basis of their resemblance and association. For instance, one of the most popular search engines “Google search engine” uses PageRank algorithm. Basically, this algorithm increases the significance of a page with the number of hyperlinks to it from other pages, as well as especially of other applicable pages. In addition, web structure mining is also widely utilized for determining “social networks” (or “community networks”) by uncovering hidden information and useful knowledge from related hyperlinks. Moreover, this term is sometimes directly linked to “link analysis” research, which is a newly emerged research area particularly designed for the use of data mining in mathematics and computer science for graph-theory and communication and social sciences for social network analysis (Lappas, 2008; Kosala & Blockeel, 2000; Cooley, Mobasher, & Srivastava, 1997). The use of Data Mining Techniques for E-Commerce Functions According to Seifert (2005), data mining entails the utilization of sophisticated data manipulation and analysis techniques and tools to find out previously unknown, relationships and valid patterns in huge data sets. In addition, these data analysis techniques and tools encompass mathematical algorithms, statistical models and machine learning techniques. However, these algorithms are intended to enhance the decision making performance automatically through experience. Also, the decision trees and neural networks are examples of this technology. Thus, data mining is a collection of various activities, such as gathering, organizing, and applying predictions and analysis of the collected data (Seifert, 2004; Piton, Blanchard, Briand, & Guillet, 2009; Turban, Leidner, McLean, & Wetherbe, 2005, p. 462). According to Yang (2006), the online marketing paradigm is becoming a speedily emerging trend. The online marketing has uniqueness as it is similar to both direct and indirect marketing. The example of direct marketing is catalog marketing and television, radio based marketing are examples of the indirect marketing. However, the capability to track web-based visits through the web based “cookies” all through a purchasing process is similar to a catalog marketer’s utilization of source codes to outline orders back to particular catalog versions. Though, the idea of advertisement simulated that is viewed through an untracked web audience is comparable to television audiences seeing a commercial. The exceptional nature of online marketing as well as the accessible data sources directs organizations to apply procedures intended for organizing and optimizing the needs of web based customers (Yang, 2006). In this scenario Seifert (2005) stated that data mining offers an automated analytical procedure for the processing of huge databases searching intended for universal patterns of customer attributes, buying behavior, contact history, or campaign feedback examination all-through millions of rows of client behavioral data. In addition, this kind of analysis techniques has been adopted by huge corporations for decades also it is a costly item. However, the clients’ data could be processed into sections and groups or segments of customers exhibited related characteristics. In addition, by applying predefined algorithms, analysis tools and automated business rules, client data and information can be processed as well as used to allocate targeted advertisement groups to offer special products or discounted rates for that specific class of web based customers. Furthermore, the purchased client “firmagraphic” or “demographic” data could also be joined to buyer records to offer extra significant information not mostly obtainable through corporation’s internal databases (Seifert, 2004; Chen & Hu, 2005). According to Thomas, Lewison, Hauser, & Foley (2006), the data mining and profiling are new and advanced techniques for the web based advertising and marketing. In addition, through these techniques, a database of prediction is built, then gathered and analyzed as much relevant information as possible to obtain the most excellent possible understanding of the target audience. For the web based advertisement Thomas, Lewison, Hauser, & Foley (2006) stated that corporations build a current customer database through the mining techniques. Then corporations build customized messages that reveal an appreciation of their clients in an extremely segmented as well as direct style (Thomas, Lewison, Hauser, & Foley, 2006). Thomas, Lewison, Hauser, & Foley (2006) outline that one of the main problems with this type of analysis (automated data mining techniques based) is that, because of the macroscopic intensity of an investigation, it is without links with the emotional procedures of a client’s interaction, for instance, preferences, self-reported attitudes as well as procedures of customer contentment. Thomas, Lewison, Hauser, & Foley (2006) stated that these procedures cannot be easily linked and integrated into the data mining procedures. Though, data mining, in its real appearance, does not present the most excellent data analysis when evaluating creative ideas. Thus, in this scenario the utilization of qualitative analytic measures is the best solution (Thomas, Lewison, Hauser, & Foley, 2006) According to Keating (2008), “the data mining is a technique to achieve market intelligence from a vast amount of data customer data”. The techniques of data mining resolve the problems of learning from business data because the lack of data is not a huge problem nowadays (Keating, 2008). Keating (2008) outline that association rules based data discovery is one of the main techniques used by the data mining in the web based advertisement and marketing. In association rules based discovery, the data mining tools find out the relations (for example relationships like what type of books definite groups of people study, what goods specific groups of people buy, what movies specific groups of people watch, etc.) among different data items. In addition, a corporation makes use of this extracted information in targeting their customers and potential marketplaces. In this scenario, there is an example of data mining tool like Netflix that recommends the movies based on movies people have viewed as well as rated in the long-ago. Furthermore, the web based selling king “Amazon” also offers similar capability in case of recommending books through viewing the customer’s buying and viewing behaviors (Keating, 2008). Vityaev & Kovalerchuk, 2008) discuss another technique used by the data mining that is clustering analysis tools. These data clustering analysis tools are extremely powerful tools intended for clustering customers and products into groups having similar nature and usage behavior. The majority of the product clusters revealed can not be utilized in company decision. Though, they can discover one or two that are enormously significant, the ones the corporations are able to take benefit of. In addition, the most widespread utilization for clustering techniques and tools is certainly in the states by economists as market segmentation. Also, in market segmentation, a corporation splits the client base into subdivisions depending upon individuality like that wealth, income, lifestyle, geographic location, etc. Furthermore, every market segment is managed and handled through a different marketing technique that effectively suited to that particular segment (Vityaev & Kovalerchuk, 2008). Li, Surendran, & Shen (2007) outline that the increase in the amount of people in the web based advertising market has produced huge volumes of data as well as stimulating data mining problems. In addition, the previous research on web pages, search logs, blogs and social network was paying attention to information association, recovery and understanding (Li, Surendran, & Shen, 2007). On the other hand, at the present, there are smart techniques for assessing the customer buying behavior and offering the most appropriate buying alternative. Recommendations Several researchers such as (Kohavi, 2001; zihao & hui, 2010) discuss the uses of data mining for e-commerce. According to their viewpoints, business firms using e-commerce for conducting business operations can to a great extent benefit from the approaches that data mining of clickstream and transactional data offer. In this scenario, these techniques can be used to renovate the web site design, customer loyalty, customization strategies and productivity. In view of the fact that the web is a wonderful and a huge experimental laboratory hence in order for the data mining to be successful and fruitful, it must consider several aspects: Huge amount of data Having a large number of records in accordance with each pattern makes sure the numerical implication of patterns has been ensured and minimized the probability of over fitting (Kohavi, 2001; zihao & hui, 2010). A number of attribute (many records) Entities which have been mined should comprise a number of attributes. In view of the fact that if the records in the data encompass simply a small number of attributes, then straightforward data mining approaches, for instance scatter plots, bar graphs and spreadsheet tables would be sufficient for effectively analyzing the data. In the same way, if this data encompasses many dozens or even hundreds of attributes, then there will be a dire need for the implementation of some automated techniques to search through the data and discover the significant elements and patterns. In addition, the research has shown the effectively designed web site can play a significant role in the collection of wide-ranging data, and a wide variety of attributes can be discovered from analysis. For instance, in case of KDD Cup 2000, more than five hundred attributes were discovered per record for the analysis (Kohavi, 2001; zihao & hui, 2010). Clean data It is an admitted fact that the presence of corrupt and noisy data and facts can put out of sight useful patterns and make it harder for the data analyst to predict. In addition, integration with legacy systems and manual data entry can bring in discrepancies and anomalies. In this scenario, straight electronic compilation of the source offers high quality and extremely dependable data. Additionally, clean data is, certainly, an ambiguous term. In this scenario, the information and data entered by humans manually for instance registration forms available on the ecommerce sites, can still have a variety of mistakes and errors however the implementation of on-line and automated confirmation system can provide an excellent support to effectively address these kinds of accidental mistakes and errors which cannot be caught in paper formats (Kohavi, 2001; zihao & hui, 2010). Actionable field Without a doubt, the implementation of these techniques can be useful for attaining useful results which are frequently discussed, on the other hand actions are hardly ever taken in actual fact for the reason that manual and traditional systems as well as nonflexible fields make it very difficult and costly to put into operation the extracted information and knowledge and enhance existing processes. In case of e-commerce web sites a variety of actions can be taken against a large number of findings by making changes to websites, comprising the site presentation, cross sells, design, up-sells, and customization. In this scenario, targeted e-mail operations can be comparatively straightforward to carry out and computerize. In addition, the analysis system and the operational system can be combined into one closed-loop system more simply than in other fields (Kohavi, 2001; zihao & hui, 2010). Computable Return on Investment (ROI) Determining changes and keeping track of their cause in ecommerce web sites is difficult, costly, and time-consuming. In addition, the web facilitates for controlled experiments (for instance use of control groups), manifold practices, and an instant assessment of their result on the objective (for instance enhanced change rates). Additionally, e-mail promotions can be tracked at every stage of the change process, which involves various steps such as opening the email, clicking through, adding products to the shopping cart, browsing products, starting an investigation, and making purchase decisions. Despite the fact that in principle e-commerce is a huge area of data mining, on the other hand, data mining is very important for eCommerce sites, social problems and business processes are up till now developing. In this scenario, business organizations must take care of channel conflicts and associated issues, reliable messages all the way through manifold touch points with customers as well as the cost required to develop and maintain electronic sites (Kohavi, 2001; zihao & hui, 2010). Conclusion Data mining has been used in a number of areas. The basic purpose of implementing data mining techniques is to extract useful information and knowledge which can be used for further analysis to improve the business processes. At the present, the use of data mining for e-commence is studied under the domain of web mining. This domain is further divided into three categories. All these categories deal with different kinds of data and knowledge extraction issues. The use of data mining tools and techniques for e-commerce brings a large number of advantages and opportunities for business organizations. This paper has presented a detailed analysis of different techniques and ways through which data mining can be used for e-commerce. This paper has discovered that, all these techniques are used to find the customer behaviors analysis, their needs analysis, segmentation, and allocation of particular e-commerce function to indentified group of customers with unique buying behaviors. This paper has also presented some recommendations that should be kept in mind while implementing data mining for e-commerce. References Chen, Y., & Hu, L. (2005). Study on data mining application in CRM system based on insurance trade . ACM International Conference Proceeding Series; Vol. 113, Proceedings of the 7th international conference on Electronic commerce (pp. 839-841). Xi'an, China : ACM New York, USA. Cooley, R., Mobasher, B., & Srivastava, J. (1997). Web mining: information and pattern discovery on the World Wide Web. 9th International Conference on Tools with Artificial Intelligence (ICTAI '97) (pp. 558-567). Newport Beach, California: IEEE. Facca, F. M., & Lanzi, P. L. (2005). Mining interesting knowledge from web logs: a survey. Data & Knowledge Engineering, Volume 5 Issue 3, pp. 225-241. Keating, B. P. (2008). Data Mining: What Is It and How Is It Used:. The Journal of Business Forecasting, pp. 1-4. Kohavi, R. (2001). Mining e-commerce data: the good, the bad, and the ugly. KDD '01 Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining (pp. 8-13). San Francisco CA USA: ACM. Kosala, R., & Blockeel, H. (2000). Web mining research: a survey. ACM SIGKDD Explorations: Newsletter of the Special Interest Group on Knowledge Discovery and Data Mining ACM, Volume 2 Issue 1, pp. 1-15. Lappas, G. (2008). An overview of web mining in societal benefit areas. Online Information Review, Volume 32 Issue 2, pp. 179-195. Laudon, K. C., & Laudon, J. P. (1999). Management Information Systems, Sixth Edition (6th ed.). New Jersey: Prentice Hall. Li, Y., Surendran, A. C., & Shen, D. (2007). Data Mining and Audience Intelligence for Advertising. ACM SIGKDD Explorations Newsletter Volume 9, Issue 2, pp. 96-99. Navarro-Arribas, G., & Torra, V. (2010). Privacy-preserving data-mining through micro-aggregation for web-based e-commerce. Internet Research, Volume 20 Issue 3, 366-384. Piton, T., Blanchard, J., Briand, H., & Guillet, F. (2009). Domain driven data mining to improve promotional campaign ROI and select marketing channels . Conference on Information and Knowledge Management, Proceeding of the 18th ACM conference on Information and knowledge management (pp. 1057-1066). Hong Kong, China: ACM New York, USA. Ranjan, J., & Bhatnagar, V. (2009). A holistic framework for mCRM – data mining perspective. Management & Computer Security, Volume 17 Issue 2, pp. 151-165. Seifert, J. W. (2004). CRS Report for Congress: Data Mining: An Overview. Congressional Research Service ˜ The Library of Congress, pp. 1-19. Ting, I.-H. (2008). Web-mining applications in e-commerce and e-services. Online Information Review, Volume 32 Issue 2, pp. 129-132. Tu, Y. (2008). An application of web-based data mining: selling strategies for online auctions. Online Information Review, Volume 32 Issue 2, pp. 147-162. Thomas, A. R., Lewison, D. M., Hauser, W. J., & Foley, L. M. (2006). Direct Marketing in Action: Cutting-Edge Strategies for Finding and Keeping the Best Customers. New York: Praeger Publishers. Turban, E., Leidner, D., McLean, E., & Wetherbe, J. (2005). Information Technology for Management: Transforming Organizations in the Digital Economy . New York: Wiley. Vityaev, E. E., & Kovalerchuk, B. Y. (2008). Relational methodology for data mining and knowledge discovery. Intelligent Data Analysis Volume 12 , Issue 2, pp. 189–210. Yang, Y. (2006). The Online Customer: New Data Mining and Marketing Approaches. Amherst, NY: Cambria Press. Zhang, Q., & Segall, R. S. (2010). Review of data, text and web mining software. Kybernetes, Volume 39 Issue 4, pp. 625-655. zihao, S., & hui, W. (2010). Research on E-commerce Application Based on Web Mining. International Conference on Intelligent Computing and Cognitive Informatics (pp. 337-340). IEEE. Read More