Projects

A. Database Usability: Processing WHY and WHY-NOT Query

After decades of efforts made by the database community, today’s systems have become highly efficient in terms of both query execution time and resource usage, and powerful in terms of provided functions. However, these systems are not usable for the end users to the same degree as they are proficient in underlying data management and query evaluation. These days, users expect systems to be more interactive and cooperative. That is, the users are not satisfied only with receiving the result from the system, but also they want to know why the system returns only the current set of objects in the result set. In particular, users may want to know why a certain data object (which is unexpected) appears in the result set, and similarly, why a certain data object (which is expected) does not appear in the result set. At present, the traditional database systems do not provide any kind of exploratory data analysis facilities for the user to support the above why and why-not questions. In this category of research, we are working on WHY and WHY-NOT Query to improve the usability of the traditional database systems.

Figure 1: A conceptual framework for answering WHY and WHY-NOT query

DP140103499 Australian Research Council Discovery Grant: On Effectively Answering Why and Why-not Questions in Databases, $373,250 for 2014-2016

Investigators: Prof C. Liu (Swinburne), Research Associate: Dr. Md. Saiful Islam

Some of the articles published already and well accepted among the database community are given as follows:

B. On Computing Top-k Promising Products: Models and Algorithms

The advancement of the World Wide Web has revolutionized the way the manufacturers can do business. The manufacturers can collect customer preferences for products and product features from their sales and other product related websites to enter and sustain in the global market. For example, the manufactures can make intelligent use of these customer preference data to decide on which products should be selected for targeted marketing in the upcoming Christmas. The selected products must attract as many customers as possible to increase the possibility of selling more than their respective competitors. In this category of research, we are investigating on product adoption model among the customers and product selection strategy among the manufacturers based on around-by semantics (e.g., dynamic and reverse skyline queries) and probability theory for discovering top-k promising products.

Product Positioning

Figure 2: Computing Top-k promising products for market sustainment

Members: Dr. M.S. Islam (Griffith), Prof C. Liu (Swinburne), Prof. W. Rahayu (La Trobe), T. Anwar (Swinburne) and Prof. Bela Stantic (Griffith)

Some of the articles published already and well accepted among the database community are given as follows:

C. Parallel Data Analytics: Indexing and Advanced Query Processing

These days, we are experiencing voluminous data everywhere such as in health care systems, industries, government agencies, computer networks including sensor and traffic networks, economics etc. Understanding this big data by discovering patterns and anomalies is crucial for improving the services provided by the government agencies and other cyber-physical systems. In this category of research, we are investigating on advanced data indexing schemes and parallel computing techniques for improving the performance of the preference-based queries involving millions of customer and product datasets.

Figure 3: QTree and Q+Tree indexing schemes for (dynamic) skyline

Members: Dr. M.S. Islam (Griffith), Prof C. Liu (Swinburne), Prof. W. Rahayu (La Trobe), T. Anwar (Swinburne) and Prof. Bela Stantic (Griffith)

Some of the articles published already and well accepted among the database community are given as follows:

D. NewNeighborhood Search in Spatial Databases: Indexing and Query Processing

The neighborhoods of a query facility are the neighborhoods that finds the query facility nearer than any other facilities. We assume here that a neighborhood can be encircled by a given radius constraint and the center of the neighborhood is always pulled towards the given facility by its special deal or the quality of services it provides to the member users of the neighborhood. That is, a neighborhood is a group of k objects encircled by radius r in a spatial domain and given a query point q, the neighborhood query retrieves the neighborhood that is nearest to q. In this project, we are working on innovative data indexing techniques which can facilitate efficient neighborhood query (and its variants) processing. Neighborhood queries have many practical applications including smart city innovation/smart urban planning where on-demand placement of a facility center such as temporary medical service center or a surveillance center for security reasons is required, which require the discovery of neighborhood instead of a single neighbor.

Figure 4: Both C1 and C2 are the neighborhoods of the query object q who find q nearer than any other facility centers f1-f6, where neighborhood is simulated by circles.

Members: Dr. M. S. Islam (Griffith), A/Prof. D. Taniar (Monash), Bojie Shen (Honours Student at Monash), Prof. W. Rahayu (La Trobe) and Naser Allheeib (PhD student at Monash)

A few recent works are given as follows:

  • M.S. Islam, D. Taniar and W. Rahayu, Efficient Processing of Reverse Nearest Neighborhood Queries on Location Data, to appear in the World Wide Web Journal, Springer, 2018. [CORE Rank: A] [2016 IF: 1.405]

  • B. Shen, M.S. Islam and D. Taniar, Direction-based Spatial Skyline for Retrieving Surrounding Objects, under review in a top-tier database conference, 2018. [CORE Rank: A]

E. NewData Exploration: No-But-Semantic-Match Query Processing

These days, keyword search queries are one of the best options for users who are not expert in submitting queries such as SQL, XPATH, and SPARQL etc. However, users often encounter failing queries (empty result) as they are unaware of the details of the underlying data source in which they issue their keyword search queries. Much of these data usability problems can be mitigated by discovering the semantic connection between the user given keywords and the existing keywords in the data source. In this category of research, we are investigating on discovering models and processing techniques for failing queries so that we can recommend interesting results and queries instead of saying “empty result” for the users.

Figure 5: A Conceptual Framework of No-BUT-Semantic-MATCH query processing Approach

Members: Prof C. Liu (Swinburne), Dr. M. S. Islam (Griffith), Dr. I. Moser (Swinburne) and M. Naseriparsa (Swinburne)

Some of the articles published already and well accepted among the database community are given as follows:

  • NewM. Naseriparsa, M.S. Islam, C. Liu and I. Moser, No-But-Semantic-Match: Computing Semantically Matched XML Keyword Search Results, to appear in World Wide Web Journal (WWWJ), 2018. [CORE Rank: A] [2016 IF: 1.405]

  • M. Naseriparsa, C. Liu, M.S. Islam and R. Zhou, XPloreRank: Exploring XML Data via You May Also Like Queries, to appear in World Wide Web Journal (WWWJ) 2018. [CORE Rank: A]

  • M. Naseriparsa, M.S. Islam and C. Liu, XSnippets: Explore Semi-Structured Data via Snippets, under review, 2018.

F. Tracking the Evolution of Congestion in Dynamic Road Networks

This project develops efficient and effective strategies for tracking the evolution of congestion in dynamic road networks. We exploit the concept of road network motifs and use the short cycles to heuristically identify the most suitable building block for an unstable road segment after transforming the road network into road graph. The research group proposes an in-memory index called Bin that compactly stores the historical sets of building blocks with no information loss and facilitates the efficient retrieval of the blocks.

Congestion Evolution Graph
Bin Index
Partitions at 7.09 AM in Melbourne

Figure 6: Tracking the evolution of congestion in dynamic road networks (a) Congestion evolution graph, (b) Bin index and (c) Partitions of congestion in Melbourne road network (07.09 AM)

Members: T. Anwar (Swinburne), Prof C. Liu (Swinburne), Dr. M. S. Islam (Griffith) and Prof. H.L. Vu (Monash).

Some of the articles published already and well accepted among the database community are given as follows:

G. A General Framework for Context-Aware Access Control to Information Resources

The computing technologies have been changing over time and in today’s dynamic environments, many organizations need to dynamically control access to information resources and services. Dr Islam is collaborating with Dr. ASM Kayes (La Trobe), Prof. Jun Han (Swinburne), Dr. Alan Colman (Swinburne) and Prof. Wenny Rahayu (La Trobe) to develop a new context-aware access control (CAAC) framework for information resources and services. The team has introduced a relationship-based access control framework by taking into account the relationship context information: types, granularity levels and strengths of the relevant relationships.

Members: Dr. ASM Kayes (La Trobe), Prof. Jun Han (Swinburne), Dr. Alan Colman (Swinburne), Dr. M.S. Islam (Griffith) and Prof. W. Rahayu (La Trobe).

Some of the articles published already and well accepted among the information systems community are given as follows:

Other works (Technical Reports):