We are recruiting Master and PhD students to work on the following projects. Please contact the project supervisors for more details.

Computer vision applications with unmanned vehicles

Dr Ivan Lee

Computer Engineering, Multimedia Systems

Suitable as PhD and Masters project

Abstract:

This project investigates computer vision application on unmanned vehicles, such as:

  1. Robot assisted smart homecare with ambient sensors: Internet of Things in smart home for detecting potential indoor accidents, 3D model reconstruction of the surrounding for identifying new objects in 3D space using a robot, human detection and pose analysis to facilitate robot-based in-situ
  2. Object detection, recognition, and tracking on a UAV: this project applies multi-camera system on a quadcopter, and algorithms for 3D model reconstruction and new object detection and tracking will be investigated in this project.

References:

  1. Kalana Withanage, Ivan Lee and Russell Brinkworth, “Mobile robotic active view planning for physiotherapy and physical exercise guidance,” IEEE International Conference on Robotics, Automation and Mechatronics (RAM), Angkor Wat, Cambodia, 2015.
  2. Victor Stamatescu, Sebastien Wong, David Kearney, Ivan Lee, and Anthony Milton, “Mutual information for enhanced feature selection in visual tracking”, SPIE Defense + Security: Automatic Target Recognition XXV, 2015.

Discovery and use of Twitter network structural features for civil unrest prediction

Professor Jiuyong Li

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Social unrest is predicable using Twitter data (Ramakrishnan et al 2014) and the structures of the Twitter networks are strong indicators. Baltimore riots and Arab Spring share many similarities in patterns of spread of messages in Twitter (Bohannon 2015). A recent study shows that there are clear network structure and community changes in Twitter after the 2011 Japanese earthquake and Tsunami (Lu and Brelsford 2014).  Another recent study in PewResearchCenter characterises six types of conversational structures in Twitters: polarized, tight crowd, Brand clusters, Community clusters, broadcast network, and support network (Smith et al 2014). This project will study the methods for extracting structural features in social networks for improving the prediction accuracy of civil unrest. Some related work for characterisation of Twitter networks can be found in (Myers 2014, Myers and Shama, 2014).

References:

  1. Ramakrishnan, N et al (2014). ‘Beating the news’ with EMBERS: forecasting civil unrest using open source indicators. KDD 2014: 1799-1808.
  2. Bohannon, J (2015). Can unrest be predicted, Science/AAAS, News May 9.
  3. Lu, X and Brelsford, C (2014). Network structure and community evolution on Twitter: human behavior change in response to the 2011 Japanese earthquake and tsunami, Nature Oct, 2014.
  4. Myers, S, Sharma, A, Gupta, P, and Lin, J (2014). The structure of the Twitter follow graph, Proceedings of International World Wide Web Conference Committee, (IW3C2 14).
  5. Myers, S, and Leskovec, J, (2014). The bursty dynamics of the Twitter information network, Proceedings of International World Wide Web Conference Committee, (IW3C2 14).

Effective time series feature selection for civil unrest prediction using social media data

Professor Jiuyong Li

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Social unrest events can be modelled by time series and are influenced by other event series (local, neighbour cities and the major cities), posts in social media, news, and economic circumstance, etc. The extracted information from the media forms features (Ramakrishnan et al 2014). Each feature is represented as a time series, and the data is a large set of time series. The aim of the project is to select a subset of time series that are informative for the prediction the future civil unrest events. Some feature selection work of time series can be found in (Kim 2012; Sun et al 2012).

References:

  1. Ramakrishnan, N et al (2014). ‘Beating the news’ with EMBERS: forecasting civil unrest using open source indicators. KDD 2014: 1799-1808.
  2. Kim, M (2012). Time-series dimensionality reduction via Granger causality. IEEE Signal Processing Letters,19(10), 611-614
  3. Sun, Y, Li, J, Liu, J, Chow, C, Sun, B, Wang, R (2014). Using causal discovery for feature selection in multivariate numerical time series, Machine Learning, advance access.

Identifying cancer subtypes from multi-levelled biological data with computational methods

Dr Thuc Le, Professor Jiuyong Li

Computer Science, Bioinformatics

Suitable as PhD and Masters project

Abstract:

Cancer is a leading cause of death, accounting for more than 8.2 million of deaths worldwide, or 22,000 people every day. In the past decade, personalised medicine, using genetic information to develop cancer-specific medication, has become a strong focus for health researchers. An important step in this personalised medicine framework is to identify cancer subtypes, as different cancer subtypes may have different treatment therapies. Since cancer is an extremely complex and heterogeneous disease, the personalised medicine framework relies heavily on achievements of advanced research in system biology (Wang, 2010).  System biology approaches use knowledge in Mathematics, Statistics and Computer Science to solve the biological problems. This project will study the computational methods for identifying cancer subtypes using multi types of biological data. Examples of related works are in (Wang et al. 2014, Liu et al. 2014). Background in Biology is an advantage but not a compulsory requirement.

References:

  1. Wang E. A roadmap of cancer systems biology. Nature Publishing Group. 2010;713: 1-28.
  2. Wang, Bo, et al. Similarity network fusion for aggregating data types on a genomic scale. Nature methods. 2014: 333-337.
  3. Liu, Yiyi, et al. “A network-assisted co-clustering algorithm to discover cancer subtypes based on gene expression.” BMC bioinformatics1 (2014): 37.

Implied Comparative Advantage of Australian Economic Complexity

Dr Ivan Lee

Computer Science, Computational Economics

Suitable as PhD and Masters project

Abstract:

This project investigates economics complexity at sub-country level in Australia (with potential extensions to global economies) based on export data within the country or to overseas, and developing new models for time series predictions of implied comparative advantage. The outcome of this project will assist policy makers identifying revealed competitive advantages and opportunity gain for different industrial sections, and predicting the industrial export growth over time. Students in this project will investigate mathematical modelling and information visualisation of economical data. This project is supported by the South Australia Department of State Development.

References:

  1. The Observatory of Economic Complexity: OEC, https://atlas.media.mit.edu/en/ (last accessed 11 June 2015)
  2. Alexander Simoes, Cesar A. Hidalgo, Juan Jimenez, Michele Coscia, Muhammed A. Yıldırım, Ricardo Hausmann, Sarah Chung, and Sebastián Bustos, “The Atlas of Economic Complexity Mapping Paths to Prosperity,” https://atlas.media.mit.edu/atlas/ (last accessed 11 June 2015)
  3. Ricardo Hausmann, Cesar A. Hidalgo, Daniel P. Stock, and Muhammed A. Yildirim, “Implied Comparative Advantage,” SSRN Electronic Journal 01/2014; DOI: 10.2139/ssrn.2410427

Integrated Prediction with Multiple Data Sources and Credibility Assessment

Dr Lin Liu

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

The research topic will focus on the problems of fusing evidence from multiple data sources and models, and the credibility of data sources and users.  The topic will be based on the work in Rekatsinas et al. (2015)’s paper on the challenge of discovering valuable sources, Hoegh et al. (2015)’s paper, in which a Bayesian model fusion framework of protest events is proposed, and the work in (Mukherjee, Weikum and Danescu-Niculescu-Mizil 2014).

Reference:

  1. Rekatsinas, T et al 2015, Finding Quality in Quantity: The Challenge of Discovering Valuable Sources for Integration. 7th Biennial Conference on Innovative Data Systems Research (CIDR‘15) January 4-7, 2015, Asilomar, California, USA
  2. Hoegh, A et al 2015, Bayesian Model Fusion for Forecasting Civil Unrest, Technometrics
  3. Mukherjee, S, Weikum,G, and Danescu-NiculescuMizil C 2014, People on drugs: credibility of user statements in health communities. In KDD’14, pages 65-74

Investigating genetic causes of cancer through complex gene regulatory networks

Dr Thuc Le, Professor Jiuyong Li

Computer Science, Bioinformatics

Suitable as PhD and Masters project

Abstract:

This project will study the computational methods for identifying the genetic causes of cancer through gene regulatory networks containing multiple gene regulators. Gene regulatory networks play an important role in every process of life, and understanding the dynamics of these networks helps reveal the mechanisms of diseases6. There have been tremendous works on inferring gene regulatory networks. However, most of the works consider the networks with only one type of gene regulator, such as transcription factors (Imam et al., 2015) or microRNAs (Le, 2013), thus only help reveal part of the whole regulatory network picture. This project aims to develop methods to construct gene regulatory networks that contain multiple types of gene regulators and methods to isolate sub-networks that are altered between normal and cancer patients. Examples of related works are in (Le et al. 2013, Ping et al. 2015). Background in Biology is an advantage but not a compulsory requirement.

References:

  1. Imam, Saheed, Daniel R. Noguera, and Timothy J. Donohue. “An Integrated Approach to Reconstructing Genome-Scale Transcriptional Regulatory Networks.” PLoS computational biology2 (2015): e1004103-e1004103.
  2. Le, Thuc D., et al. “Inferring microRNA–mRNA causal regulatory relationships from expression data.” Bioinformatics6 (2013): 765-771.
  3. Le, Thuc D., et al. “Inferring microRNA and transcription factor regulatory networks in heterogeneous data.” BMC bioinformatics1 (2013): 92.
  4. Ping, Yanyan, et al. Identifying core gene modules in glioblastoma based on multilayer factor-mediated dysfunctional regulatory networks through integrating multi-dimensional genomic data. Nucleic acids research4 (2015): 1997-2007.

Multimedia Systems (2D and 3D video coding and video streaming, robotics vision, cloud-based video services, panoramic video analysis, video surveillance and monitoring, multimedia data mining, multimedia sensor networks, medical imaging)

Dr Ivan Lee

Computer Engineering

Suitable as PhD and Masters project

Abstract:

Multimedia systems use a combination of content forms to facilitate media rich applications such as video conferencing or robotics vision. The multimedia projects we offer include either software or hardware design, developing applications for mobile devices (smart phone, tablets), desktop computers, robots, embedded systems, or high-performance computers (such as clusters or cloud computers.) The candidates will have opportunities to utilise different sensors, such as 2D and 3D video cameras, microphone arrays, marker/visual-based tracking systems, or the Australian Synchrotron, for different projects.

Potential projects include, but not limited to:

  • 2D and 3D video coding
  • Compressive video coding
  • Free-viewpoint video coding and streaming
  • Cloud-based video streaming
  • Vision system for unmanned aerial vehicle (UAV), Unmanned ground vehicle (UGV), or autonomous underwater vehicle (AUV)
  • Wireless multimedia sensor networks
  • Medical imaging
  • Biomechanics using computer vision

Precursor Pattern Analysis and Interpretable Classification

Dr Lin Liu

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

The aim is to develop new pattern discovery methods on large-scale unstructured online data to derive useful relationships among instances of variables, especially targeted at those messages prior to the civil unrest events so that interpretable prediction models with causal relationships can be learned based on the methods developed in (Letham et al. 2013; Li, Liu and Le 2015).

References:

  1. Letham, B et al 2013, Interpretable classifiers using rules and Bayesian analysis: Building a better stroke prediction model. Technical Report no. 609, University of Washington, August 2013.
  2. Li,J, Liu,L, and Le, T 2015, Practical approaches to causal relationship exploration, Springer, 2015

Predicting Dengue Virus Infections using Weibo Data

Dr Jixue Liu

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Weibo is the Chinese Twitter where people use Chinese to discuss their opinions, feelings, and symptoms when they fall ill. This project aims to explore how to extract information from Chinese Weibo and how to build models for predicting dengue outbreaks in China. The built models then become a comparison to the one built for Australia. The relevance of this PhD project is that when China has an outbreak, the risk of outbreak in Australia is increased because of the large number of travellers between the two countries.

Reference:

  1. Rekatsinas, T., et al. (2015). “SourceSeer: Forecasting Rare Disease Outbreaks Using Multiple Data Sources.” SIAM International Conference on Data Mining (SDM).
  2. Gomide, J., et al. (2011). “Dengue surveillance based on a computational model of spatio-temporal locality of Twitter.” ACM Web Science Conference (WebSci).
  3. Chan, E. H., et al. (2011). “Using Web Search Query Data to Monitor Dengue Epidemics: A New Model for Neglected Tropical Disease Surveillance.” PLOS Neglected Tropical Diseases.
  4. Sang, S., et al. (2014). “Predicting Local Dengue Transmission in Guangzhou, China, through the Influence of Imported Cases, Mosquito Density and Climate Variability.” PLOS ONE 9(7).
  5. Kim, E.-K., et al. (2013). “Use of Hangeul Twitter to Track and Predict Human Influenza Infection.” PLOS ONE.

Relationship of Dengue Infections in Different Cities in Twitter Data

Dr Jixue Liu

Computer Science

Suitable as PhD and Masters project

Abstract:

This project aims to model the relationship between dengue outbreaks in different Australia cities or/and in cities of a neighboring country like Malaysia. Twitter data and government disease statistics will be used in prediction. The model is then used to predict the outbreak of a specific city based on the infections of other cities. This model will be useful if it is adapted to cities across cities in different countries.

Reference:

  1. Rekatsinas, T., et al. (2015). “SourceSeer: Forecasting Rare Disease Outbreaks Using Multiple Data Sources.” SIAM International Conference on Data Mining (SDM).
  2. Gomide, J., et al. (2011). “Dengue surveillance based on a computational model of spatio-temporal locality of Twitter.” ACM Web Science Conference (WebSci).
  3. Chan, E. H., et al. (2011). “Using Web Search Query Data to Monitor Dengue Epidemics: A New Model for Neglected Tropical Disease Surveillance.” PLOS Neglected Tropical Diseases.
  4. Hirose, H. and L. Wang (2012). “Prediction of Infectious Disease Spread Using Twitter: A Case of Influenza.” Fifth International Symposium on Parallel Architectures, Algorithms and Programming (PAAP): 100-105.

Signal processing and analysis for medical imaging

Dr Ivan Lee

Computer Engineering, Multimedia Systems

Suitable as PhD and Masters project

Abstract:

This project will investigate sparse signal reconstruction for computed tomography and particle image velocimetry analysis, on synchrotron phase contract x-ray images, to overcome challenges on detecting and tracking overlapping particles for the assessment of cystic fibrosis airway therapies. This project can also apply similar algorithm for different medical imaging techniques, such as MRI, ultrasound, and confocal microscopic images.

References:

  1. Zhenglin Wang and Ivan Lee, “Backprojection Regularization with Weighted Ramp Filter for Tomographic Reconstruction”, International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Milan, Italy, 2015.
  2. Hyewon Jung, Ivan Lee, Sang-Heon Lee, “Circular Particle Detection using Sectored Ring Mask for Synchrotron PCXI images,” International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Milan, Italy, 2015.

 

INTEGRATED POLICING: Generating queries for identity resolution

Dr Jixue Liu

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Identity resolution aims to find whether two records are referring to the same entity. A lot of work has been done on this topic assuming that the databases containing the two records to be matched are fully accessible and enabling brute force comparison of all records. However, this full access assumption becomes impractical in some applications because a database may be too sensitive to be accessed in any way that a user likes to take. When such a database is accessed, the user may get only one record per query or get even only true/false answers. In this case, what queries should be used to access the database so that an identity resolution process can use the query output to infer identities becomes a serious problem. This project aims to develop methods and algorithms to generate best queries to access the restricted database for identity resolution purpose. The query generation would be on the basis of an existing index over similar entities, e.g., the similar names.

References:

  1. Heng Ji. 2015. From Mono-lingual to Cross-lingual: state-of-the-art EDL. Invited Talk at JHU HLT-COE
  2. Heng Ji, Joel Nothman and Ben Hachey. 2014. Overview of TAC-KBP2014 Entity Discovery and Linking Tasks. Proc. Text Analysis Conference (TAC2014)
  3. Dan Roth, Heng Ji, Ming-Wei Chang and Taylor Cassidy. 2014. Wikification and Beyond: The Challenges of Entity and Concept Grounding. Tutorial at the 52nd Annual Meeting of the Association for Computational Linguistics (ACL2014)
  4. Roy et al (2005). “Towards Automatic Association of Relevant Unstructured Content with Structured Query Results.” CIKM.
  5. Gardezi et al (2012). “Query Rewriting using Datalog for Duplicate Resolution.” LNCS 7494: 86-98.
  6. Talburt, J., Entity and Identity Resolution. MIT IQ Industry Symposium http://mitiq.mit.edu/IQIS/2010/Addenda/T2A%20-%20JohnTalburt.pdf, 2010.
  7. Christen, P., A Survey of Indexing Techniques for Scalable Record Linkage and Deduplication. IEEE Transactions on Knowledge and Data Engineering (TKDE), 2012. 24(9): p. 1537-1555.

 

INTEGRATED POLICING: Model relationships from text data for identity resolution

Dr Jixue Liu

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Identity resolution aims to find whether two records are referring to the same entity. When identity resolution is required from entities in text documents, the task becomes complicated. One reason is that a documents often refers to many entities and properties of entities (like names of people) are not labelled by attributes. At the same time, the relationships among the entities are described in the documents. For example, a document may contain the sentence ‘Alice saw that Bob drove down Stephen St at 11:00pm’. Here three names are mentioned alongside of the time entity and the relationships between these names are described. Currently, the methods dealing with this type of text use Natural Language Parsing tools to extract entities and then put the entities into relational tables and the match with other relational records of a database.  The shortage of this practice is that the relationships are not used. This project aims to model the entities and the relationships extracted in text data, and develop ways to compare the modelled entities and relationships with records in relational databases. The model is expected to be a graph model. The comparison will need an effective method and an efficient algorithm.

References:

  1. Xiang Ren, Ahmed El-Kishky, Heng Ji and Jiawei Han. Automatic Entity Recognition and Typing in Massive Text Data. Tutorial at ACM International Conference on Management of Data (SIGMOD2016)
  2. Heng Ji, Joel Nothman and Ben Hachey. 2015. Overview of TAC-KBP2015 Tri-lingual Entity Discovery and Linking. Proc. Text Analysis Conference (TAC2015)
  3. Gardezi et al (2012). “Query Rewriting using Datalog for Duplicate Resolution.” LNCS 7494: 86-98.
  4. Bruce, J., et al., Pathways To Identity: Using Visualization To Aid Law Enforcement In Identification Tasks. Security Informatics, 2014. 3(12).
  5. Xu, J., et al., Complex Problem Solving: Identity Matching Based on Social Contextual Information. Journal of the Association for Information Systems, 2007. 8(10): p. 525-545.
  6. Christen, P., A Survey of Indexing Techniques for Scalable Record Linkage and Deduplication. IEEE Transactions on Knowledge and Data Engineering (TKDE), 2012. 24(9): p. 1537-1555.

 

Efficient Causal Inference in Big Data

Dr Kui Yu, Professor Jiuyong Li

Computer Science

Suitable as PhD and Masters project

Abstract:

Causal inference is a fundamental problem in science. The access to big data has opened up new opportunities for inferring causal relationships from purely observational data when experimental tests and interventions are difficult or unethical. Most of existing causal discovery algorithms are designed for a small and single data set. Thus, big data brings great challenges on causal inference because of its volume, the diversity of data types and the speed at which it must be managed. The project will develop efficient and effective causal inference algorithms to deal with big data challenges for advancing big data mining techniques, and extend those new algorithms to discover genetic causes of cancer for improving biomedical discovery. The novel causal inference methods developed in the project will advance data mining techniques and help human being better understanding cause-and-effect relationships hidden in big data. By extending the research outcomes to discover genetic causes of cancer to help biomedical researchers understand critical causes and trends buried in big biomedical data, this will bring great potential to improve biomedical discovery for better healthcare in Australia.

References:

  1. Liang and A. R. Mikler. (2014) Big data problems on discovering and analyzing causal relationships in  pidemiological data. IEEE BigData 2014, 11-18.
  2. Yu, W. Ding, H. Wang, and X. Wu. (2013) Bridging Causal Relevance and Pattern Discriminability: Mining Emerging Patterns from High-Dimensional Data. IEEE Transactions on Knowledge and Data Engineering, 25(12): 2721-2739.
  3. F. Cooper, I. Bahar, M. J. Becich, P. V. Benos and et.al. (2015) The center for causal discovery of biomedical knowledge from Big Data, Journal of the American Medical Informatics Association, 1-6.​

Developing novel data mining techniques for mining educational data

Professor Jiuyong Li, Dr Lin Liu

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

This project aims to develop data mining techniques for effectively identifying the factors that influence student academic performance and building better models to predict student learning outcomes. The increasing adoption of learning management systems, such as Moodle has enabled education institutions to collect a large of amount of data related to student online activities. Findings from such data can assist the institutions to provide timely and effective student support and to make interventions. Educational data mining [1] has been attracting more and more research interests in recent years. However, due to the large volume and high complexity of the data logged by the learning management systems, traditional data mining methods are facing new challenges to deal with the big educational data to find out true influential factors on student performance and to build accurate and interpretable models to predict student outcomes. This project will develop new methods, such causal discovery approaches [2] to tackle the educational data mining challenges.

References:

  1. Cristobal Romero, Sebastian Ventura, and Enrique Garcıa. Data mining in course management systems: Moodle case study and tutorial. Computers & Education, 51 (1). pp. 368-384, 2008
  2. Jiuyong Li, Lin Liu, and Thuc Le. Practical Approaches to Causal Relationship Exploration. Springer, 2015.

 

Interpretable classification and prediction of civil unrest events

Professor Jiuyong Li, Dr Jie Chen

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

The lack of interpretability makes many sophisticated classification/regression models infeasible in real applications, which place great emphasis on both the accuracy and comprehensibility of the potential models, such as medical scoring systems (Letham et al. 2015). The goal of the project is to build interpretable prediction models with patterns, which are easy for human reasoning and understanding.  There is recent progress on Bayesian analysis (Letham et al. 2015) and discriminative pattern-based classification (Shang et al. 2016; Lou et al. 2013).  The patterns or high-order features that are highly correlated to the target civil unrest events will be used as the input of the building of the models.  The student may compare the performance of the two major interpretable models in the context of predicting civil unrest events, and explore a new way of interpretable prediction model, e.g. through the combination the two approaches.

References:

  1. Letham, B. et al (2015). “Interpretable classifiers using rules and Bayesian analysis: Building a better stroke prediction model.” The Annals of Applied Statistics 9.3 (2015): 1350-1371.
  2. Shang, J., et al. (2016). An Effective but Concise Discriminative Patterns-Based Classification Framework. In Proceedings of 2016 SIAM International Conference on Data Mining (SDM 2016)
  3. Lou, Y. et al (2013). Accurate intelligible models with pairwise interactions. In SIGKDD, 2013, 623–631

Automatic labelling of tweets in civil unrest prediction

Professor Jiuyong Li, Dr Wei Kang

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Twitter data is considered as an important open source when predicting civil unrest events. A number of models have been built with features/patterns extracted from tweets, such as the volume-based model and planned protest model in (Ramakrishnan et al. 2014), and the forward-looking approach to crowd behaviour prediction in (Kallus 2014). However, labelling of tweets still remains a challenging task due to the nature of tweets. In (Zhao et al. 2014), the authors manually labelled 5386 tweets as civil unrest related, and 6147 as unrelated, which required a large amount of labor force. The objective of this project is to design and implement either unsupervised or semi-supervised approaches(Hua et al. 2013), so as to label tweets automatically.

References:

  1. Ramakrishnan, N., Butler, P., Muthiah, S., Self, N., Khandpur, R., Saraf, P., Wang, W., Cadena, J., Vullikanti, A., Korkmaz, G., others, 2014. “Beating the news” with EMBERS: forecasting civil unrest using open source indicators, in: Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, pp. 1799–1808.
  2. Kallus, N., 2014, April. Predicting crowd behavior with big public data. InProceedings of the companion publication of the 23rd international conference on World wide web companion (pp. 625-630). International World Wide Web Conferences Steering Committee.
  3. Zhao, L., Chen, F., Dai, J., Hua, T., Lu, C.T. and Ramakrishnan, N., 2014. Unsupervised spatial event detection in targeted domains with applications to civil unrest modeling. PloS one, 9(10), p.e110206.
  4. Hua, T., Chen, F., Zhao, L., Lu, C.-T. & Ramakrishnan, N. STED: semi-supervised targeted-interest event detection.  KDD’13, 1466-1469.

 

Prediction of civil unrest events with news and other data sources

Professor Jiuyong Li, Dr Jie Chen

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

Tweets has played important role in the prediction of civil unrest events, such as protests, and may provide insights into the root causes of the events.  However, online news feeds, blogs and other sources e.g. economic time series and GDELT data, are also useful in the forecasting of these events (Ramakrishnan et al. 2014). The informative patterns discovered from the news sources, e.g. interactive patterns (Ning et al.2015) and precursor patterns (Ning eg al. 2016), can be utilised in enhancement of existing predictive models that majorly rely on twitter data.

References:

  1. Ramakrishnan N., Butler P., Muthiah S, et al. “Beating the news” with EMBERS: forecasting civil unrest using open source indicators. KDD ′14, New York, ACM, August24–27, 2014 pp. 1799–1808
  2. Yue Ning, Sathappan Muthiah, Ravi Tandon, Naren Ramakrishnan: Uncovering News-Twitter Reciprocity via Interaction Patterns. ASONAM 2015: 1-8
  3. Yue Ning, Sathappan Muthiah, Huzefa Rangwala, Naren Ramakrishnan: Modeling Precursors for Event Forecasting via Nested Multi-Instance Learning. CoRR abs/1602.08033 (2016)

Integration and visualisation of multiple civil unrest prediction models

Professor Jiuyong Li, Dr Wei Kang

Computer Science, Data Mining

Suitable as PhD and Masters project

Abstract:

A complete system often consists of multiple models/components, which work and interact with each other to provide expected results. The objective of this project is to integrate multiple existing civil unrest prediction models (Ramakrishnan et al 2014) into one system, and make sure all the components work properly together to provide a comprehensive and user-friendly result through user interface and visualisation.

References:

  1. Ramakrishnan, N., Butler, P., Muthiah, S., Self, N., Khandpur, R., Saraf, P., Wang, W., Cadena, J., Vullikanti, A., Korkmaz, G., others, 2014. “Beating the news” with EMBERS: forecasting civil unrest using open source indicators, in: Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, pp. 1799–1808.