دانلود مقاله ISI انگلیسی شماره 3417
عنوان فارسی مقاله

ادغام نظریه رزونانس تطبیقی شبکه عصبی و الگوریتم ژنتیک K- ابزار برای تجزیه و تحلیل مسیرهای مرور وب در تجارت الکترونیک

کد مقاله سال انتشار مقاله انگلیسی ترجمه فارسی تعداد کلمات
3417 2005 20 صفحه PDF سفارش دهید 6910 کلمه
خرید مقاله
پس از پرداخت، فوراً می توانید مقاله را دانلود فرمایید.
عنوان انگلیسی
Integration of ART2 neural network and genetic K-means algorithm for analyzing Web browsing paths in electronic commerce
منبع

Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)

Journal : Decision Support Systems, Volume 40, Issue 2, August 2005, Pages 355–374

کلمات کلیدی
2 - تجزیه و تحلیل خوشه ای - داده کاوی -     نظریه رزونانس تطبیقی - الگوریتم ژنتیک - ابزار - سیستم عامل توصیه -
پیش نمایش مقاله
پیش نمایش مقاله ادغام نظریه رزونانس تطبیقی شبکه عصبی و الگوریتم ژنتیک K- ابزار برای تجزیه و تحلیل مسیرهای مرور وب در تجارت الکترونیک

چکیده انگلیسی

Neural networks and genetic algorithms are useful for clustering analysis in data mining. Artificial neural networks (ANNs) and genetic algorithms (GAs) have been applied in many areas with very promising results. Thus, this study uses adaptive resonance theory 2 (ART2) neural network to determine an initial solution, and then applies genetic K-means algorithm (GKA) to find the final solution for analyzing Web browsing paths in electronic commerce (EC). The proposed method is compared with ART2 followed by K-means. In order to verify the proposed method, data from a Monte Carlo Simulation are used. The simulation results show that the ART2+GKA is significantly better than the ART2+K-means, both for mean within cluster variations and misclassification rate. A real-world problem, a recommendation agent system for a Web PDA company, is investigated. In this system, the browsing paths are used for clustering in order to analyze the browsing preferences of customers. These results also show that, based on the mean within-cluster variations, ART2+GKA is much more effective.

مقدمه انگلیسی

Electronic commerce (EC) has developed rapidly in recent years. Because entering the Internet is not difficult and creating customer interaction is easy on the Internet, how to create long-term customer relationships is a critical factor for successful EC. The Institute for Information Industry (III) showed that there were over six million Internet users by the end of 2000, and it has increased dramatically. This results in more requirements for the analysis of network loading and more complexity of Web site design. To help users in browsing Web contents is an important factor for designing a Web site. Thus, browsing behavior becomes an important index of effectiveness of a site. By analyzing the frequency of Web page clicks and by understanding the regular browsing paths, the structure of Web site can be improved and more popular Web pages can be provided to the customers, thereby increasing EC sales. Therefore, this research proposes a novel clustering analysis technique for data mining. It is employed to analyze the browsing paths and behaviors of EC customers in order to improve the development of Web sites and customer satisfaction. Clustering analysis is a common tool in multivariate analysis and has been applied widely in many areas. The purpose of clustering analysis is to determine the objects in the same cluster with similar characteristics. Furthermore, it determines when there is a significant difference between two different clusters. The applications of clustering analysis include social science, genetics, biology, business and education. In addition to statistical methods, artificial neural networks (ANNs) have also been widely applied in such areas. The unsupervised neural network, which is able to cluster objects by learning from training samples, is especially useful. Genetic algorithms (GAs) also have this capability. Kuo et al. [11] have reported that very good solutions can be provided by using self-organizing feature maps (SOM) of neural networks to determine the number of clusters and the starting points, and then employing the K-means method to find the final solution. They also showed that K-means can be replaced by GA in order to get better results [11]. [10] Thus, this current study proposes using a modified two-stage clustering method, adaptive resonance theory 2 (ART2) neural network [3] and [4], followed by genetic K-means algorithm (GKA). The results are compared with those from ART2 followed by K-means. In order to evaluate the performance of the two clustering analysis methods, both simulation and real-world data are employed. Both of sets of results indicate that the proposed method is better than ART2 followed by K-means. Then, the proposed method is applied to create a recommendation agent system for a PDA Web company, which can dramatically increase customer satisfaction. The remainder of this paper is organized as follows. Section 2 introduces some necessary background, including applications of neural networks and genetic algorithms for clustering analysis. The proposed method is presented in Section 3, and Section 4 illustrates the simulation results. Section 5 shows the model evaluation results for a PDA Web company, and concluding remarks are made in Section 6.

نتیجه گیری انگلیسی

In a clustering problem, it is always difficult to determine the number of clusters. Therefore, this study proposes a two-stage method, which first uses the Adaptive Resonance Theory 2 (ART2) to determine the number of clusters and an initial solution, then using genetic K-means algorithm (GKA) to find the final solution. Sometimes, the self-organizing feature map with two-dimensional output topology has great difficulty determining the number of clusters by observing the map. However, ART2 can actually determine the number of clusters according to the number of output nodes. Through Monte Carlo simulation and a real case problem, the proposed two-stage clustering analysis method, ART2+GKA, has been shown to provide high performance. The p-value of Scheffe's multiple comparison test shows that the two cluster methods, ART2+K-means and ART2+GKA, do not differ significantly, but the average of within cluster variance of ART2+GKA is less than that of ART2+K-means. This may be because ART2+GKA has the characteristics of both a genetic algorithm and K-means. In the real-world case study, the browsing paths of a Web PDA company were clustered by both ART2+K-means and ART2+GKA. The results from ART2+GKA are much better than those of ART2+K-means, demonstrating that ART+GKA is an efficient tool for clustering analysis. Based on the clustering result, the Web administrators can make more effective Webs for the customers. After understanding the customers' behaviors, the recommendation mechanism can be more easily and precisely created. Because customers in the same cluster have similar characteristics, this mechanism can recommend Web pages that are clicked more frequently for the same cluster. Thus, for a larger Web site, this process may save search time, thus increasing the customers' convenience and satisfaction. Future research can improve the current study. For instance, the parameters of GKA, like the crossover and mutation rates, affect the time to converge, so an experimental design can decide the best parameter combination to speed up the convergence. Because an ant colony system is also effective at searching, it may be a good candidate for replacing the ART2 neural network.

خرید مقاله
پس از پرداخت، فوراً می توانید مقاله را دانلود فرمایید.