Towards evaluating and enhancing the reach of online health forums for smoking cessation (Link)
- Abstract Online pro-health social networks facilitating smoking cessation through web-assisted interventions have flourished in the past decade. In order to properly evaluate and increase the impact of this form of treatment on society, one needs to understand and be able to quantify its reach, as defined within the widely adopted RE-AIM framework. In the online communication context, user engagement is an integral component of reach. This paper quantitatively studies the effect of engagement on the users of the Alt.Support.Stop-Smoking forum that served the needs of an online smoking cessation community for more than 10 years. The paper then demonstrates how online service evaluation and planning by social network analysts can be applied towards strategic interventions targeting increased user engagement in online health forums. To this end, the challenges and opportunities are identified in the development of thread recommendation systems for effective and efficient spread of healthy behaviors, in particular smoking cessation.
Low Cost Sensor System Design for In-home Physical Activity Tracking (Link)
- Abstract An aging and more sedentary population requires interventions aimed at monitoring physical activity, particularly within the home. This research uses simulation, optimization, and regression analyses to assess the feasibility of using a small number of sensors to track movement and infer physical activity levels of older adults. Based on activity data from the American Time Use Survey and assisted living apartment layouts, we determined that using 2-3 doorway sensors can be used to effectively capture a sufficient amount of movements in order to estimate activity. The research also identified preferred approaches for assigning sensor locations, evaluated the error magnitude inherent in the approach, and developed a methodology to identify which apartment layouts would be best suited for these technologies.
Sequential Stochastic Assignment with Unknown Worker Quality (Master’s thesis | Link)
- Abstract The Sequential Stochastic Assignment Problem (SSAP), introduced by Derman in 1972, lies in anticipatively assigning workers with given success rates to sequentially-arriving tasks in order to maximize the total expected reward; every task value follows a given distribution and remains unknown until the task presents itself to the workers. This work studies a real-world extension of this problem in which the quality of the workers is not known prior to making assignments. Such a problem setting has practical significance in multiple real-world instances owing to the fact that the true success rate of a worker is seldom easily obtained. We use techniques for inferring the ranking of the workers and provide a policy that dictates the assignment of workers to the incoming tasks. The inference techniques in question are borrowed from existing work in the literature, one of which uses a random-walk based approach to ranking-inference and another that uses an L1 – based heuristic to infer ranking information. We use the idea behind the random walk based inference to incorporate two methods – node based and permutation based random walk SSAP – to devise rankings and employ them with the standard SSAP results. We also compare these three methods against the case where complete information about worker rankings is known. Although the SSAP with complete information about worker rankings provides us with the highest expected reward, we note that the node based random walk SSAP is better in the absence of complete information.
Network Design for Transporting Perishable Edible Commodities with the Application of Queuing Theory (Bachelor’s Thesis | Link)
- The supply of agricultural products has received a great deal of attention lately due to issues related to public health and degradation of commodity quality. Something that has become apparent is that in the near future the design of supply chain mechanism will become subject to more stringent regulations and monitoring, in particular those for products destined for human consumption. This work is a study for determining the impact of transportation time and shelf life of the Tea, commodity being considered, on the structure of the transportation network connecting tea plantations to retail units via warehouses using road freight in Assam. The objective of the problem is to minimize the transportation cost of the system and the warehousing facilities in a way that the total transportation time remains significantly lower than the shelf life of the commodity. The problem is solved by using a cutting plane method that incorporates cuts generated through a set of tangent hyper-planes extended from the convolute of the perfectly convex waiting time function at the intermediate warehouses. A C++ code that implements CPLEX concert libraries (Version 12.2) has been used to generate numerical results based on random data sets to check functionality of proposed model. The effects of the transportation cost, transportation speed, and of the shelf life of the tea being transported on the fixed cost and the total network operational cost of the network have been further discussed.
Reaction Paper : Influence Maximization in Social Networks (Link)
- Description: Influence Maximization is an extremely hot topic in the field of Social Network Analysis and involves studying how influence spreads through a Social Network given that it started from a specific node or a set of nodes. This reaction paper looks at some of the biggest contributions made within the topic on Influence Maximization. While it originally began as a project proposal, I believe that this reaction paper provides a pretty good insight into the state of the art and criticisms.
Review of Estimating and Understanding Exponential Random Graph Models (Link)
- Description: Exponential Random Graph Models (ERGM) are a family of statistical models for analyzing data about social and other networks. The underlying principle of ERGMs is said to be embedded within the concepts of Gibbs entropy. This paper by Chatterjee, analyses ERGMs from the point of view of the Principle of Large Deviations and provides proofs on obtaining the standard ERGM results. You can view the original paper by Chatterjee here
Graph Similarity on a MapReduce Framework (Link)
- Description: Pairwise comparisons are often useful in situations where there is a need to perform multi-criteria decision making or to choose a strategy/item from a set of alternatives. When the strategies/items can be formulated as a relational databases or Graphs, the problem becomes exponentially hard and one is forced to rely on heuristics. This class project report addresses the computational intractability of performing pairwise comparisons over a large number of graphs. The widely used MapReduce framework is implemented to cut-down on the computational cost and running time involved in performing a pairwise analysis over a Graph dataset. The results are compared against a traditional sequential algorithm to show a significant improvement in run-times.