Project Title: Applying Data Provenance to the domain of Scientific Publication
- Collaborators: Dr. Tariq Mahmood, Dr. Imran Jami, Dr. Zubair Shaikh, Mr. Hussain Mughal
- Time Period: March 2011 – February 2012
- Summary: The concept of "provenance" refers to the documented history of a digital artifact. A common digital artifact is a scientific publication (research paper), which results from the collaborative efforts of a group of authors (scientists). We implemented a provenance-aware system for documenting publications, called PADS. It employs a three-layered provenance hierarchy, which provides provenance information for the process of (from general to specific): 1) managing the logistics related to a research paper (submission, acceptance, registration etc.), 2) documenting the research paper, i.e., how different sections of this publication are linked together, and 3) linking the research contents (e.g., limitations, idea, etc.) between different sections of this paper. This hierarchy can output useful provenance information, e.g., the conferences having a given acceptance rate, the sections of a paper typically documented by a given author, the research contents available in a particular section etc. This provenance information can also be used to generate three different provenance models, or profiles: 1) profile for research ventures (conferences, workshops etc.), 2) profile for reviewers, and 3) profile for authors. We perform some experiments with PADS and validate its performance.
Project Title: Learning Dynamically Personalized Strategies in Conversational Recommender Systems
- Collaborators: Dr. Tariq Mahmood, Ghulam Mujtaba (IBA-Sukkur)
- Time Period: January 2010 – January 2012
- Summary: Conversational Recommender Systems (CRSs) are intelligent E-Commerce applications which interactively assist online users to acquire their interaction goals during their sessions. In this paper, we develop and implement a methodology that allows CRSs to autonomously learn and update an optimal strategy dynamically (at run-time), and individually for each separate user. We validate our approach in an off-line simulation with four simulated users, who interact with a travel CRS, as well as in an online evaluation with thirteen real users, who interact with a prototype for a book CRS. Notwithstanding the computational inefficiency of RL in a dynamic setting, we show that our approach is able to dynamically learn and update the optimal strategy for each real and simulated user. For the simulated users, the optimal behavior is reasonably adapted to the users’ characteristics. For the real users, the optimal behavior is also adapted insofar that it provides assistance in certain situations, allowing many users to buy several books together.
Project Title: Prediction of Terrorist Events in Pakistan to Support Counter-terrorism
- Participants: Dr. Tariq Mahmood, Shafaq Khan
- Summary: This project uses Artificial Neural Networks to make a predictive model for different types of terrorist events in Pakistan. We have identified 9 different types of predictive configurations which can provide very useful results to counter-terrorism authorities in thwarting terrorist attacks.
Project Title: Adaptive Automated Teller Machines (AATMs)
- Participants: Dr. Tariq Mahmood, Ghulam Mujtaba
- Summary: In this project, we have mined the ATM transactions of bank customers. We have discovered that withdrawl is the most common operation, followed by balance inquiry and then purchase. We have also mined the statistical processes of frequent bank transactions. Based on our results, we have proposed a set of adaptive ATM interfaces that can be used to minimize the ATM usage time, along with providing a better ATM experience for the users.
Project Title: Process Mining Subscription Requests of Users to Telecommunication Services
- Participants: Dr. Tariq Mahmood, Arsalan Ahmed
- Summary: In this project, we have applied process mining on the subscription requests of users to different telecommunication services of an anonymous company, e.g., Breaking News, Namaz Alerts etc. The results reveal the different time periods of increased subscription requests, for different services, along with the generic and probabilistic model of requests. These results have enabled business managers of the concerned company to increase their profits.
Project Title: Detecting Communities in Twitter Networks By Mining Tweets
- Participants: Dr. Tariq Mahmood, Munir Hussain
- Summary: In this project, we apply clustering techniques on several different types of Twitter profiles. The results tell us that more similarity is found between users who belong to some Twitter community, e.g., Data Mining
Project Title: Volunteer-based Emergency and Charity Portal
- Participants: Dr. Tariq Mahmood, Javaria Mansoor
- Summary: We have implemented a mobile application that allows registered volunteers to quickly reach an emergency location, within a given radius of their registered address. The application also allows charity to be distributed to the appropriate people who have volunteered for it.