Nabeel Abdur Rehman
Research Associate at ITU, Punjab 2013-current
Research Assistant at LUMS 2011-2013
Bachelors in Electrical Engineering from LUMS 2009-2013
Teaching Assistant Artificial Intelligence course at LUMS Fall 2013
Teaching Assistant Development Lab 2 course at ITU Fall 2013

This page has not been updated since: 7 April 2014. For latest updates please visit my website
Link to my website (link)


About Me

A recent graduate from LUMS majored in Electrical Engineering, currently working as a Research Associate at NEWT Lab.
My research interests include large scale Data Analysis, Technology for Developing Countries, Health Informatics and Social Media Analysis. Currently, I am working on various research projects with Dr. Umar Saif and Dr. Lakshminarayanan Subramanian.


Current Research Projects

Fine-grained dengue surveillance with citizen-driven data

Currently, I am working on a citizen-driven early-warning system using data from a phone helpline system that was set up in collaboration with the Punjab province government. The outbreak prediction system is modeled as a phone-based analog of Google Flu Trends. where phone calls to the helpline are used as a proxy to predict the intensity of a suspected outbreak at the town-level upto 14 days ahead. There are two advantages of using phone calls data over the search query data. Firstly, in developing countries like Pakistan, people often have no access to internet, thus search queries from such areas are not a useful indicator of the disease activity in the area. In contrast, telephones and cellphones are owned by most people and data from this source highly correlates with actual disease activity. Secondly, the phone calls data is available at a town-level. This means that disease activity can be monitored on a very fine-granularity level as small as 1/10th of a city level. Developing countries like Pakistan where budgets are very constrained, targeted containment can be performed using such system.
We presented the early findings of this work along with the other projects we are doing with Punjab Government in the mHealth Summit. Washington DC, Dec 2012.
This project is in near completion and we are planning to send our findings in a high impact venue.

Preventing Dengue outbreaks in Developing Countries

In 2011, Pakistan faced its deadliest Dengue outbreak affecting over 16000 people. Amongst the various steps Punjab government took to contain future outbreaks, one initiative was to use android phones to geo-tag containment activities. The initiative turned out to be very successful in containing dengue outbreaks in 2012 and 2013.
In this project we are analyzing, how various containment activities performed by workers helped in containing the dengue outbreak. We are analyzing the spatial and temporal effects of individual containment activities on reducing the number of dengue cases. We are also analyzing how such systems can help reduce corruption and how such low cost systems can be replicated in any other developing country to control disease outbreaks.
This project is also in near completion and we are planning to send our findings in a high impact venue.

How interesting is Twitter for disease surveillance?

I am also working on a project of analyzing the usefulness of twitter data for disease surveillance. Several recent works on Twitter-based disease surveillance have relied on the basic assumption that user tweets provide significant amount of information about disease related issues in a given location. In this project we aim to ask the inverse question: "Is Twitter indeed useful as a data source for disease surveillance in a given location?". To analyze this question, we have extracted a corpus of around 2 million from over a period of 14 months about two popular infectious diseases: dengue and flu. Using available location and language information available in a subset of tweets, we have classified the data into different buckets and analyzed each bucket in isolation across a host of standard tweet metrics used in disease surveillance.


Completed Research Projects

FluBreaks

I have worked in developing an Early Epidemic Detection System which uses Google Flu trends data as its basis. The Google Flu Trends service was launched in 2008 to track changes in the volume of online search queries related to flu-like symptoms. Over the last few years, the trend data produced by this service has shown a consistent relationship with the actual number of flu reports collected by the US Centers for Disease Control and Prevention (CDC), often identifying increases in flu cases weeks in advance of CDC records. Our system uses a wide range of algorithms to monitors the changes in the trends data and generates outbreak alerts on detecting an anomaly. We compared the accuracy and practicality of three types of algorithms: normal distribution algorithms, Poisson distribution algorithms, and negative binomial distribution algorithms. We explored the relative merits of these methods, and related our findings to changes in Internet penetration and population size for the regions in Google Flu Trends providing data. My major contribution to this work was the design and implementation of outbreak detection algorithms and interpretation of results. We published our findings in Journal of Medical Internet Research (JMIR).
Read our paper (link)
Link to our system (link)

Identifying Trending Topics in Urdu short messages

SMSAll, a very popular short messaging service (SMS) in Pakistan, is used by over 2.7 million people to broadcast Urdu messages written in Roman English to manually created groups. Identification of trending topics in the SMSAll dataset is a challenging problem as the short messages broadcast by users often do not follow any grammatical structure and Romanizing Urdu words in English language means that there is no fixed spelling for a single word. In this project, we developed a supervised learning method using a language agnostic model for identification of trending topics in the SMSAll dataset. The language agnostic model was based on a novel idea of connecting words. Using this model we were able to identify topics in the dataset with a very high accuracy.

Dengue Early Epidemic Detection System for Punjab Government

I have also been involved in developing an early epidemic detection system for Punjab government to monitor Dengue Disease. The system uses actual geo-tagged cases data and positive larva data to generate alerts. The system comprises of a combination of spatio-temporal and purely temporal algorithms and outputs alerts in two forms. 1) Center point and radius of high alert areas. 2) High alert UCs(UCs are the smallest unit of area covering approx. 1000-5000 houses). We used Google Maps API for easy visualization of the alerts. The system is currently being used by government officials to direct workers to perform targeted containment activities to curb dengue outbreaks at earlier stage.
Link to our system (link)

Characterizing Dengue Spread and Severity using Internet Media Sources

I have also worked in building a system that automatically aims to characterize the spread and severity of the dengue disease at a fine-grained location granularity based on analyzing news reports from Internet media sources. Our system leverages a range of standard data mining and machine learning techniques to arrive at an accurate dengue severity measure for any given location. Based on a detailed analysis of news reports gathered from several leading dailies in Pakistan, we demonstrated the effectiveness of our system to accurately characterize the dengue spread and severity across different locations within Pakistan. My major contribution to this work was to employ various machine learning techniques and algorithms to obtain the most relevant features extracted from the news articles and use them to classify the spread and severity of dengue disease. Our finding were accepted as a poster in Proceedings of the 3rd ACM Symposium on Computing for Development(DEV).
Read our paper (link)
Currently, we are working on extending this system to cover Malaria disease in Pakistan.


Publications

  • Fahad Pervaiz, Mansoor Pervaiz, Nabeel Abdur Rehman, Umar Saif, "FluBreaks: Early Epidemic Detection from Google Flu Trends", Journal of Medical Internet Research 2012;14(5):e125 (link)

  • Talal Ahmad, Nabeel Abdur Rehman, Fahad Pervaiz, Shankar Kalyanaraman, Maaz Bin Safeer Ahmad, Sunandan Chakraborty, Umar Saif, Lakshminarayanan Subramanian. "Characterizing Dengue Spread and Severity using Internet Media Sources", Proceedings of the 3rd ACM Symposium on Computing for Development(DEV) (link)

  • Fahad Pervaiz, Talal Ahmad, Nabeel Abdur Rehman, Umar Saif, Lakshminarayanan Subramanian. “Punjab-IDSS: Dengue Surveillance, Early Detection and Containment 2012”, mHealth Summit. Washington DC, Dec 2012.





Contact

Neighbourhood for Emerging World Technologies (NEWT)
Information Technology University
Arfa Technology Park
Lahore, Pakistan
Email: nabeel(dot)abdur(dot)rehman(at)gmail(dot)com;
          nabeel(dot)rehman(at)itu(dot)edu(dot)pk