Jae-Young Chang

Query reformulation using materialized views in data warehouse environment

Materialized views offer opportunities for significant performance gain in query evaluation by providing fast access to pi-e-computed data. The question of when and how to use a materialized view in processing a given query is a difficult one attracting a significant amount of research. In previous works, only materialized views whose relations are contained in those of a query have been used and, as a result, certain potentially useful materialized views were excluded from consideration. Proposed in this paper are new ways of utilizing materialized views in answering a query with aggregation operations; Views including relations not referred to in the given query are utilized. We identify the conditions where a materialized view can be used in reformulating a query. Also presented are algorithms to find the most efficient reformulated query. The proposed conditions and corresponding algorithms provide significant and practical performance improvements to the data warehousing environment.

An extended query reformulation technique using materialized views

Materialized views offer opportunities for significant performance gain in query evaluation by providing fast access to pre-computed data. The question of when and how to use a materialized view in processing a given query is a difficult one attracting a significant amount of research. In previous works only one-to-one or containment mapping from views to a query has been used and, as a result, certain potentially useful materialized views were excluded from consideration. Proposed in this paper are new ways of utilizing materialized views in answering a query. Views including relations not referred to in an original query, which were excluded in previous works, are utilized. Attributes missing from a view can be recovered under certain conditions. We present the conditions where a view may be used in these ways and algorithms that can effectively test these conditions and reformulate the query. The proposed conditions and corresponding algorithms provide a significant and practical extension to the usability of materialized views in query processing.

Intelligent Data Prefetching for Hybrid Flash-Disk Storage Using Sequential Pattern Mining Technique

This paper presents an intelligent prefetching technique that significantly improves hybrid flash-disk storage, a combination of hard disk and flash memory. As a prefetching strategy, we adopt the sequential pattern mining, a variant of association rule mining. Our goal is to minimize overall I/O processing time of hybrid storage systems with using the Fully Associated Sector Translation (FAST) technique that is known to be the best mapping method in managing flash memory. It is very significant to further enhance the system performance of the hybrid storage when applying FAST to it. In our work, the hybrid storage uses the flash memory as a cache space to improve system performance. With this memory architecture, the proposed method is to prefetch objects onto ‘prefetching’ blocks in the level of both file and block in hybrid storage systems. Through extensive experiments using real UCC data and synthetic data, we show that the proposed prefetching method outperforms conventional ones.

Integrating Incremental Feature Weighting into NaÏve Bayes Text Classifier

In the real-world operational environment, text classification systems should handle the problem of incomplete training set and no prior knowledge of feature space. In this regard, the most appropriate algorithm for operational text classification is the naive Bayes since it is easy to incrementally update its pre-learned classification model and feature space. Our work mainly focuses on improving naive Bayes classifier through feature weighting strategy. The basic idea is that parameter estimation of naive Bayes can consider the degree of feature importance as well as feature distribution. In addition, we have extended a conventional algorithm for incremental feature update for developing a dynamic feature space in operational environment. Through experiments using the Reuters-21578 and the 20 Newsgroup benchmark collections, we show that the traditional multinomial naive Bayes classifier can be significantly improved by chi2-statistic based feature weighting.

An Experimental Evaluation of Short Opinion Document Classification Using A Word Pattern Frequency

An opinion mining technique which was developed from document classification in area of data mining now becomes a common interest in domestic as well as international industries. The core of opinion mining is to decide precisely whether an opinion document is a positive or negative one. Although many related approaches have been previously proposed, a classification accuracy was not satisfiable enough to applying them in practical applications. A opinion documents written in Korean are not easy to determine a polarity automatically because they often include various and ungrammatical words in expressing subjective opinions. Proposed in this paper is a new approach of classification of opinion documents, which considers only a frequency of word patterns and excludes the grammatical factors as much as possible. In proposed method, we express a document into a bag of words and then apply a learning algorithm using a frequency of word patterns, and finally decide the polarity of the document using a score function. Additionally, we also present the experiment results for evaluating the accuracy of the proposed method.

Zoned-partitioning of tree-like access methods

The performance of access methods and the underlying disk system is a significant factor in determining the performance of database applications, especially with large sets of data. While modern hard disks are manufactured with multiple physical zones, where seek times and data transfer rates vary significantly across the zones, there has been little consideration of this important disk characteristic in designing access methods (indexing schemes). Instead, conventional access methods have been developed based on a traditional disk model that comes with many simplifying assumptions such as an average seek time and a single data transfer rate. The paper proposes novel partitioning techniques that can be applied to any tree-like access methods, both dynamic and static, fully utilizing zoning characteristics of hard disks. The index pages are allocated to disk zones in such a way that more frequently accessed index pages are stored in a faster disk zone. On top of the zoned data placement, a localized query processing technique is proposed to significantly improve the query performance by reducing page retrieval times from the hard disk.

Automatic Retrieval of SNS Opinion Document Using Machine Learning Technique

Recently, as Social Network Services(SNS) are becoming more popular, much research has been doing on analyzing public opinions from SNS. One of the most important tasks for solving such a problem is to separate opinion(subjective) documents from others(e.g. objective documents) in SNS. In this paper, we propose a new method of retrieving the opinion documents from Twitter. The reason why it is not easy to search or classify the opinion documents in Twitter is due to a lack of publicly available Twitter documents for training. To tackle the problem, at first, we build a machine-learned model for sentiment classification using the external documents similar to Twitter, and then modify the model to separate the opinion documents from Twitter. Experimental results show that proposed method can be applied successfully in opinion classification.

An optimization of disjunctive queries: union-pushdown

Most previous works on query optimization techniques deal with conjunctive queries only because the queries with disjunctive predicates are complex to optimize. Hence, for disjunctive queries, query optimizers based on these techniques generate plans using rather simple methods such as CNF- and DNF-based optimization. However, the plans generated by these methods perform extremely poorly for certain types of queries. The authors propose new query optimization method, union-pushdown, for disjunctive queries. This method is composed of four phases, and each phase utilizes some advantageous techniques of CNF- and DNF-based methods. They analyze the performance of the union-pushdown plan against those of conventional plans and show that union-pushdown can be applied to various disjunctive query types without performance degradation.

Improving Naïve Bayes Text Classifier with Modified EM Algorithm

This paper presents the method of significantly improving conventional Bayesian statistical text classifier by incorporating accelerated EM (Expectation Maximization) algorithm. EM algorithm experiences a slow convergence and performance degrade in its iterative process, especially when real textual documents do not follow EM’s assumptions. We propose a new accelerated EM algorithm that is simple yet has a fast convergence speed and allow to estimate a more accurate classification model on Bayesian text classifier.

Extended conditions for answering an aggregate query using materialized views

A view is a virtual relation defined in terms of ba se relations and is said to be materialized if it is stored in the database. In the past few years, mate rialized views have attracted a significant amount of research in many applications environments as a means of enhancing query performance. Materialized views offer significant performance ad vantages in evaluating a query by eliminating the need for recomputing the views. Just as a cache , a materialized view is a type of replicated copy of information derived from base relations and, hen ce, the technique is very useful in various applications where access to local or cached views may be cheaper than access to the base relations. Although many approaches for utilizing materialized views in evaluating a query have been proposed [1, 2, 4, 5], there were several restricti ons in selecting such views. First, only views whose relations are contained in those of a query w ere considered. Thus, if a view refers to relations not mentioned in the query, it was excluded from consideration. Furthermore, a view is also excluded if it does not contain the necessary attributes for the query. However, there are many cases where a view can be utilized even if the abov e restrictions are not satisfied. In this paper, we propose a new approach to using m aterialized views in answering an aggregate query. We extend the practical scope of utilizing m aterialized views which include those that would have been excluded in previous approaches. We first show the conditions for testing whether a materialized view can be utilized in answering a qu ery even if the materialized view has relations not mentioned in the query. These conditions are then extended by introducing a method that recovers the missing attributes. All conditions are designed to be applicable for ba g (multiset) semantics. Bag semantics are used in the underlying models of most practical database management systems.