Algorithms and Applications for Academic Search, Recommendation and Quantitative Association Rule Mining

Description

Algorithms and Applications for Academic Search, Recommendation and Quantitative Association Rule Mining presents novel algorithms for academic search, recommendation and association rule mining that have been developed and optimized for different commercial as well as academic purpose systems. Along with the design and implementation of algorithms, a major part of the work presented in the book involves the development of new systems both for commercial as well as for academic use. In the first part of the book the author introduces a novel hierarchical heuristic scheme for re-ranking academic publications retrieved from standard digital libraries. The scheme is based on the hierarchical combination of a custom implementation of the term frequency heuristic, a time-depreciated citation score and a graph-theoretic computed score that relates the paper's index terms with each other. In order to evaluate the performance of the introduced algorithms, a meta-search engine has been designed and developed that submits user queries to standard digital repositories of academic publications and re-ranks the top-n results using the introduced hierarchical heuristic scheme. In the second part of the book the design of novel recommendation algorithms with application in different types of e-commerce systems are described. The newly introduced algorithms are a part of a developed Movie Recommendation system, the first such system to be commercially deployed in Greece by a major Triple Play services provider. The initial version of the system uses a novel hybrid recommender (user, item and content based) and provides daily recommendations to all active subscribers of the provider (currently more than 30,000). The recommenders that we are presenting are hybrid by nature, using an ensemble configuration of different content, user as well as item-based recommenders in order to provide more accurate recommendation results. The final part of the book presents the design of a quantitative association rule mining algorithm. Quantitative association rules refer to a special type of association rules of the form that antecedent implies consequent consisting of a set of numerical or quantitative attributes. The introduced mining algorithm processes a specific number of user histories in order to generate a set of association rules with a minimally required support and confidence value. The generated rules show strong relationships that exist between the consequent and the antecedent of each rule, representing different items that have been consumed at specific price levels. This research book will be of appeal to researchers, graduate students, professionals, engineers and computer programmers. Cover Half Title Series Page Title Page Copyright Page Table of Contents Abstract Acknowledgements List of Figures List of Tables 1 1.1 Introduction 1.2 Algorithmic Motivation and Objectives 1.3 Related Work 1.4 Algorithmic Challenges 2: Academic Search Algorithms 2.1 Collecting Data from Scientific Publications 2.2 Topic Similarity Using Graphs 2.2.1 Graph Construction 2.2.2 Type I Graph 2.2.3 Type II Graph 2.3 Topic Similarity Using Graphs 2.4 System Architecture 2.5 Heuristic Hierarchy 2.5.1 Term Frequency Heuristic 2.5.2 Depreciated Citation Count Heuristic 2.5.3 Maximal Weighted Cliques Heuristic 2.6 Experiments’ Design 2.7 Experimental Results 2.7.1 Comparisons with ACM Portal 2.7.2 Comparison with other Heuristic Configurations 2.7.3 Comparison with Other Academic Search Engines 2.7.4 Can PubSearch Promote Good Publications “Buried” in ACM Portal Results? 2.7.5 Run-Time Overhead 3: Recommender Systems 3.1 System Architecture Overview 3.1.1 AMORE Web Service 3.1.2 AMORE Batch Process 3.2 Recommender Ensemble 3.2.1 Recommendation Approach 3.2.2 Content-Based Recommender 3.2.3 Item-Based Recommender 3.2.4 User-Based Recommender 3.2.5 Final Hybrid Parallel Recommender Ensemble 3.2.6 Experiments with Other Base Recommender Algorithms 3.3 Computational Results 3.4 User and System Interfaces 4: Quantitative Association Rules Mining 4.1 Why Quantitative Association Rules? 4.2 Algorithm Overview 4.3 Algorithm Design 4.4 Recommender Post-Processor 4.4.1 Overview 4.4.2 Post-Processing Algorithm 4.5 Synthetic Dataset Generator 4.6 Configuration Parameters 4.7 Item Demand Elasticity 4.8 Dataset Generation Process 4.8.1 Generation Cycle 4.8.2 Update Cycle 4.9 Experimental Results 4.9.1 Metric 4.9.2 QARM Results Using Synthetically Generated Datasets 4.9.3 QARM Results Using Movielens Dataset 4.9.4 QARM Results Using Post-Processor 5: Conclusions and Future Directions References Index About the Author