Juggapong Natwichai

Hiding classification rules for data sharing with privacy preservation

In this paper, we propose a method of hiding sensitive classification rules from data mining algorithms for categorical datasets. Our approach is to reconstruct a dataset according to the classification rules that have been checked and agreed by the data owner for releasing to data sharing. Unlike the other heuristic modification approaches, firstly, our method classifies a given dataset. Subsequently, a set of classification rules is shown to the data owner to identify the sensitive rules that should be hidden. After that we build a new decision tree that is constituted only non-sensitive rules. Finally, a new dataset is reconstructed. Our experiments show that the sensitive rules can be hidden completely on the reconstructed datasets. While non-sensitive rules are still able to discovered without any side effect. Moreover, our method can also preserve high usability of reconstructed datasets.

A Novel Heuristic Algorithm for Privacy Preserving of Associative Classification

Since individual data are being collected everywhere in the era of data explosion, privacy preserving has become a necessity for any data mining task. Therefore, data transformation to ensure privacy preservation is needed. Meanwhile, the transformed data must have quality to be used in the intended data mining task, i.e. the impact on the data quality with regard to the data mining task must be minimized. However, the data transformation problem to preserve the data privacy while minimizing the impact has been proven as an NP-hard. In this paper, we address the problem of maintaining the data quality in the scenarios which the transformed data will be used to build associative classification models. We propose a novel heuristic algorithm to preserve the privacy and maintain the data quality. Our heuristic is guided by the classification correction rate (CCR) of the given datasets. Our proposed algorithm is validated by experiments. From the experiments, the results show that the proposed algorithm is not only efficient, but also highly effective.

Data Quality in Privacy Preservation for Associative Classification

Privacy preserving has become an essential process for any data mining task. In general, data transformation is needed to ensure privacy preservation. Once the privacy is preserved, data quality issue must be addressed, i.e. the impact on data quality should be minimized. In this paper, k-Anonymization is considered as the transformation approach for preserving data privacy. In such a context, we discuss the metrics of the data quality in terms of classification, which is one of the most important tasks in data mining. Since different type of classification may use different approach to deliver knowledge, data quality metric for the classification task should be tailored to a certain type of classification. Specifically, we propose a frequency-based data quality metric to represent the data quality of the transformed dataset in the situation that associative classification is to be processed. Subsequently, we validate our proposed metric with experiments. The experiment results have shown that our proposed metric can effectively reflect the data quality for the associative classification problem.

Knowledge maintenance on data streams with concept drifting

Concept drifting in data streams often occurs unpredictably at any time. Currently many classification mining algorithms deal with this problem by using an incremental learning approach or ensemble classifiers approach. However, both of them can not make a prediction at any time exactly. In this paper, we propose a novel strategy for the maintenance of knowledge. Our approach stores and maintains knowledge in ambiguous decision table with current statistical indicators. With our disambiguation algorithm, a decision tree without any time problem can be synthesized on the fly efficiently. Our experiment results have shown that the accuracy rate of our approach is higher and smoother than other approaches. So, our algorithm is demonstrated to be a real anytime approach.

Workflow-Based Composite Job Scheduling for Decentralized Distributed Systems

The development of distributed systems has been accelerated in this recent years through the development of P2P, grid, and cloud computing. Nevertheless, job scheduling is still a major challenge for distributed system design and implementation. Such issue can be more complex when dealing with workflow-based composite jobs, i.e. each job has multiple tasks with dependencies between them. In this paper we prove that, when task dependency in such kind of jobs is pre-defined with workflow templates, finding minimal execution time for multiple jobs is an NP-complete problem. So, we propose a heuristic algorithm for this problem. The experimental results show that this algorithm can find near optimal solutions to the problem with a polynomial computation time.

Privacy preservation for associative classification: an approximation algorithm

Privacy is one of the most important issues when dealing with the individual data. Typically, given a data set and a data-processing target, the privacy can be guaranteed based on the pre-specified standard by applying privacy data-transformation algorithms. Also, the utility of the data set must be considered while the transformation takes place. However, the data-transformation problem such that a privacy standard must be satisfied and the impact on the data utility must be minimised is an NP-hard problem. In this paper, we propose an approximation algorithm for the data transformation problem. The focused data processing addressed in this paper is classification using association rule, or associative classification. The proposed algorithm can transform the given data sets with O(k log k)-approximation factor with regard to the data utility comparing with the optimal solutions. The experiment results show that the algorithm is both effective and efficient comparing with the optimal algorithm and the other two heuristic algorithms.

Incremental privacy preservation for associative classification

Privacy preserving has become an essential process for any data mining task. Therefore, data transformation to ensure privacy preservation is needed. In this paper, we address a problem of privacy preserving on an incremental-data scenario in which the data need to be transformed are not static, but appended all the time. Our work is based on a well-known data privacy model, i.e. k-Anonymity. Meanwhile the data mining task to be applied to the given dataset is associative classification. As the problem of privacy preserving for data mining has proven as an NP-hard, we propose to study the characteristics of a proven heuristic algorithm in the incremental scenarios theoretically. Subsequently, we propose a few observations which lead to the techniques to reduce the computational complexity for the problem setting in which the outputs remains the same. In addition, we propose a simple algorithm, which is at most as efficient as the polynomial-time heuristic algorithm in the worst case, for the problem.

Content-based video search on peer-to-peer networks

In this paper, we address the efficiency issue of the content-based video indexing over peer-to-peer (P2P) networks. The traditional video index is improved to suit with the P2P computational model. The algorithms to perform the video query based on the content similarity in the P2P environment are proposed. Also, the algorithms to handle the node joining, departure, index entry insertion are proposed. Furthermore, the load balancing approach based on the proposed algorithms is proposed. From the experiment results, our proposed approach outperforms a naive approach, which directly applies the P2P model with full replication, when a number of P2P nodes to be joined, as well as a number of videos to be inserted, is increased. Meanwhile, the efficiency of our approach in terms of the query answering is bounded by linear complexity. Moreover, our proposed load balancing approach is much more efficient than the naive approach in all experiments.

A Heuristic Algorithm for Workflow-Based Job Scheduling in Decentralized Distributed Systems with Heterogeneous Resources

Decentralized distributed systems, such as grids, clouds or networks of sensors, have been widely investigated recently. An important nature of such systems is the heterogeneity of their resources; in order to archive the availability, scalability and flexibility. As a consequence, managing the systems to meet requirements is obviously a nontrivial work. The issue is even more challenging in term of job scheduling when the task dependency within each job exists. In this paper, we address such problem of job scheduling, so called workflow-based job scheduling, in the decentralized distributed systems with heterogeneous resources. As such problem is proven to be an NP-complete problem, an efficient heuristic algorithm to address this problem is proposed. The algorithm is based on an observation that the heterogeneity of the resources can affect the execution time of the scheduling. We compare the effectiveness and efficiency of the proposed algorithm with a baseline algorithm. The result shows that our algorithm is highly effective and efficient for the scheduling problem in the decentralized distributed system with heterogeneous resources environment both in terms of the solution quality and the execution time respectively.

An Efficient Indexing for Top-k Query Answering in Location-Based Recommendation System

Location-based recommendation systems are obtaining interests from both the business and research communities recently. In this paper, we propose an indexing approach to improve the efficiency of the top-k query answering for a prominent location-based recommendation model, User-centered collaborative location and activity filtering (UCLAF). The efficiency issue of such query type is important since there could be enormous users, locations, and activities in the recommendation model. When a query is issued, not all of the answers are to be obtained, but only a few most-relevant answers. Our proposed work is based on a multi-dimensional index, aR-Tree. A feature of such index tree, i.e. only-relevant information traversal, is utilized with some modification. In addition, the experiments have been conducted to evaluate our proposed work. In which, the results, which our work is compared with a few indexing methods, show that our work is highly efficient when system is scaled up.