Giovanni Giuffrida

A Fast Algorithm for Hierarchical Text Classification

Text classification is becoming more important with the proliferation of the Internet and the huge amount of data it transfers. We present an efficient algorithm for text classification using hierarchical classifiers based on a concept hierarchy. The simple TFIDF classifier is chosen to train sample data and to classify other new data. Despite its simplicity, results of experiments on Web pages and TV closed captions demonstrate high classification accuracy. Application of feature subset selection techniques improves the performance. Our algorithm is computationally efficient being bounded by O(n log n) for n samples.

A scalable bottom-up data mining algorithm for relational databases

Machine learning induction algorithms are difficult to scale to very large databases because of their memory-bound nature. Using virtual memory results in a significant performance degradation. To overcome such shortcomings, we developed a classification rule induction algorithm for relational databases. Our algorithm uses a bottom-up rule generation strategy that is more effective for mining databases having large cardinality of nominal variables. We have successfully used our algorithm to mine a retail grocery database containing more than 1.6 million records in about 5 hours on a dual Pentium processor PC.

Intelligent information dissemination services in hybrid satellite-wireless networks

The need for rapid deployment and user mobility suggest the use of a hybrid satellite‐wireless network infrastructure for important situation awareness and emergency response applications. An Intelligent Information Dissemination Service (IIDS) has been developed to support the dissemination and maintenance of extended situation awareness throughout such a network information infrastructure in a seamless manner. One of the goals of IIDS is to transparently handle the mismatches in characteristics of satellite and terrestrial wireless networks, allow effective utilization of available bandwidth, and support timely delivery of highly relevant information. IIDS achieves the above by implementing user profile aggregation that incrementally aggregates users into communities sharing common interests to enable multicast‐based information dissemination. Based on the user grouping, semantic profile matching customizes information streams based on matching user group interest profiles. By taking into account expected changes in user profiles, profile‐oriented data dissemination achieves predictive push and caching that anticipates future user needs and minimizes latency of data request by making data available before they are explicitly requested. Finally, bandwidth‐aware filtering adapts information streams to resource bandwidth availability to gracefully hide the bandwidth mismatch between the satellite and wireless links in the hybrid network infrastructure. The IIDS software has been deployed on the Digital Wireless Battlefield Network (DWBN) that integrates commercial off‐the‐shelf satellite and wireless products into a heterogeneous satellite/wireless hybrid network for supporting wireless mobile multimedia services.