Search and Classification of "Interesting" Business Applications in the World Wide Web using a Neural Network Approach

Karl Kurbel, Kirti Singh, Frank Teuteberg
Europe University Viadrina Frankfurt (Oder), Germany


A database of business Internet applications (Internet Database, IDB for short) has been developed at Europe University Viadrina in Frankfurt (Oder), Germany. The purpose of IDB is to document the state-of-the-art of business Internet use, with a focus on business-to-business applications. The database was described before in (Kurbel, 1997).

The number and the diversity of WWW applications are huge and still growing fast. As we are primarily interested in innovative solutions for the database, the question is how to find those solutions. One way is, of course, "manual" search, i.e. using a keyword-based search engine (e.g. Alta Vista) or going through company listings in the WWW ("yellow pages", business servers, etc.) and analyzing the respective WWW offers. This is obviously a tiring and time-consuming task. Automation of the search process is desirable.

The term WWW offer is used to describe the general fact that a company offers specific information for business partners via the Internet. The approach proposed in this paper is to have an automated process pre-select WWW offers which might be of interest for the database. The final decision is made by a human. The pre-selection step is based on a neural network approach. It reduces the information overload in the Web by categorizing WWW offers. For this purpose, 15 versions of the multi-layer perceptron with error back-propagation (Wassermann, 1989), four versions of generalized feed-forward networks and four versions of modular networks were trained and tested.

Since the focus of the database is on business-to-business, consumer-oriented WWW offers are considered "not interesting". In the business-to-business field, "interesting" applications are applications where business deals are substantially based on Internet use. More precisely, two classification schemes presented before are employed to distinguish between the two groups of "interesting" and "not interesting" WWW offers. In (Kurbel, 1997), business-to-business applications were categorized under two points of view: a) direct communication between business partners, and b) communication with or via information exchanges.

For the first group (direct communication), six main categories were defined:

  1. Providing information
  2. Providing information plus contact offer
  3. Starting a transaction via Internet
  4. Starting and completing a transaction via Internet
  5. Business-process interfaces via Internet

Categories 1) and 2) are typically the ones where a company brochure is put in the Web but the reader cannot do much more than just read it. WWW offers of those categories are declared "not interesting" whereas offers of categories 3) to 6) have the potential of being "interesting" for IDB.

For the second group (information exchanges), four categories were proposed (Kurbel, 1997):

  1. Simple catalog of firms and/or products
  2. Catalog with search option
  3. Product exchange
  4. Electronic market

Here the categories 1) and 2) are regarded as the not interesting ones. Categories 3) and 4) contain those offers where the mediator operating the information exchange gives automated support for transactions between business partners. Those categories are considered "interesting".

In the next section, the neural networks used for the classification task are introduced. The subsequent section then describes the process of automated search and the classification approach. Afterwards, the specific configurations of the networks are presented and experimental results are discussed. A summary and an outlook to future improvements are given in the concluding section.

To read this full paper please click the "Download Document" button below:

Looking for predictive data analytics and neural network software to solve your own problem? Try NeuroSolutions Infinity free for 14-days! NeuroSolutions Infinity is the easiest, most powerful neural network software of the NeuroSolutions family. It streamlines the data mining process by automatically cleaning and preprocessing your data. Then it uses distributed computing, advanced neural networks, and artificial intelligence (AI) to model your data.