الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 144 |  |  |
| | | from | 144 | | |
Abstract
Recommender systems (RSs) have proven to be valuable means for online users to cope with the information overload and have become one of the most powerful and popular tools in electronic commerce. RSs are software tools and techniques providing available suggestions of items/persons to the user, hence, they typically apply techniques and methodologies from Data Mining. The most frequently used technique is the classification as it matches the aims of RSs that basically classify items based on user’s preferences. However most of the common RSs are designed for a specific domain. This thesis introduces an Intelligent Adaptive Vertical Recommendation (IAVR) system that can be adapted to work in any domain. IAVR consists of four layers, which are ; (i) Content Analyzer Layer (CAL), (ii) Profile Learning Layer (PLL), (iii) Collaborative Layer (CL), and (iv) Matching and Ranking Layer (MRL). Six classification techniques are adopted to enhance the performance of the IAVR system. These techniques are ; Neuro Fuzzy (NF) classifier, Modified K-Nearest Neighbor (MKNN) classifier, Merged Binary Classifier (MBC), Accumulative Naïve Bayes (ANB) classifier, Association rule based Classifier (AC), and Merged Multi-class Classifier (MMC). The performance of the NF, MKNN, and MBC classifiers is investigated. The MBC classifier is found to be the best one that can be use in the distiller module to elect documents related to the Domain Of Interest (DOI). While the MMC classifier is used in the Multi Classification Module (MCM) as it is found more accurate than the ANB and AC classifiers. This classifier is used to classify those related documents to one of the predefined domain hypotheses. The MMC classifier is the only one used in the Profile Learning Layer. This classifier helps in discovering the user’s area of interest in order to build his profile. For this purpose the thesis introduces a profile learning model. Experimental results have proven the effectiveness of the proposed CAL and PLL layers of the proposed IAVR system with its classifiers, which accordingly promote the recommendation accuracy of the whole system. It should be noted that both CL and MRL layers are out of scope this thesis.