This paper presents an immune-based model for Web data mining, in which the rule library and statistic dictionary are used to build search schemas, and data from different sources are wrapped to XML. By using Web content mining and Web structure mining, data are abstracted to mediated schemas and then compared with search schemas. As the information about Web page content and its links are optimized, the traditional search strategy based on keyword matching can be improved. At the same time, many ideas of immune system are introduced to increase the search efficiency. Now many Web data mining system using immune approaches by Andrew Seeker, Alex A. Freitas, and Jon Timmis, (2003), H.A. Abbass, R.A. Sarker, and C.S. Newton (2001) are concerned with data clustering, while immunity is mainly used for data processing after clustering in this model.

Several new results which contribute to the understanding of parallel merging networks are presented. First, a simple new explanation of the operation of Batcher's merging networks is offered. This view leads to the derivation of a modified version of Batcher's odd-even (m, n) network which has delay time [log(m+n)]. This is the same delay time as Batcher's bitonic (m, n) network, but it is achieved with substantially fewer comparators. Second, a correspondence is demonstrated between the number of comparators (and the delay time) for such networks and certain properties of binary number systems which have recently been extensively studied. Third, the [log(m + n)] delay time is shown to be optimal for a non-degenerate range of values of m and n.