Krishna Madhavan

Mining Social Media Data for Understanding Students’ Learning Experiences

Students informal conversations on social media (e.g., Twitter, Facebook) shed light into their educational experiences-opinions, feelings, and concerns about the learning process. Data from such uninstrumented environments can provide valuable knowledge to inform student learning. Analyzing such data, however, can be challenging. The complexity of students experiences reflected from social media content requires human interpretation. However, the growing scale of data demands automatic data analysis techniques. In this paper, we developed a workflow to integrate both qualitative analysis and large-scale data mining techniques. We focused on engineering students Twitter posts to understand issues and problems in their educational experiences. We first conducted a qualitative analysis on samples taken from about 25,000 tweets related to engineering students college life. We found engineering students encounter problems such as heavy study load, lack of social engagement, and sleep deprivation. Based on these results, we implemented a multi-label classification algorithm to classify tweets reflecting students problems. We then used the algorithm to train a detector of student problems from about 35,000 tweets streamed at the geo-location of Purdue University. This work, for the first time, presents a methodology and results that show how informal social media data can provide insights into students experiences.

Models for early prediction of at-risk students in a course using standards-based grading

Using predictive modeling methods, it is possible to identify at-risk students early and inform both the instructors and the students. While some universities have started to use standards-based grading, which has educational advantages over common score-based grading, at-risk prediction models have not been adapted to reap the benefits of standards-based grading in courses that utilize this grading. In this paper, we compare predictive methods to identify at-risk students in a course that used standards-based grading. Only in-semester performance data that were available to the course instructors were used in the prediction methods. When identifying at-risk students, it is important to minimize false negative (i.e., type II) error while not increasing false positive (i.e., type I) error significantly. To increase the generalizability of the models and accuracy of the predictions, we used a feature selection method to reduce the number of variables used in each model. The Naive Bayes Classifier model and an Ensemble model using a sequence of models (i.e., Support Vector Machine, K-Nearest Neighbors, and Naive Bayes Classifier) had the best results among the seven tested modeling methods. Seven different prediction models for identifying at-risk students were compared.Only in-semester performance factors (i.e., grades) were used in the models.Models were created based on standards-based grading.Feature selection method resulted in higher accuracy of the models.

A Web-Based Tool for Collaborative Social Media Data Analysis

User-generated content on social media sites such as Twitter and Facebook provides opportunity for researchers in various fields to understand human behaviors and social phenomena. On the one hand, these human behaviors and social phenomena are very complex in nature thus require in-depth qualitative analysis. On the other, the magnitude of social media data requires large-scale data analysis techniques. In this paper, we propose a web-based tool named SWAB (Social Web Analysis Buddy) that integrates both qualitative analysis and large-scale data mining techniques. Specifically, this tool supports asynchronous collaboration among researchers conducting inductive content analysis on textural data from users online posts and conversations. It then aggregates the results and calculates the agreement among researchers, and builds modeling algorithms based on the qualitative results to classify large-scale social media text content. This current paper focuses on the overall workflow and user interface design of this tool. We demonstrate the prototype of this tool by analyzing student-posted content on Twitter.

Designing for STEM Faculty: The Use of Personas for Evaluating and Improving Design

We demonstrate in a case study how we used a qualitative method for user modeling, persona, to evaluate and refine the design of an interactive visual analytics tool. We explain the literature and our methods for building the persona, and its used to (a) evaluate existing design decisions; (b) make new design decisions in order to serve a specific user group, faculty in STEM. We present the results of 24 in-depth, qualitative interviews with STEM faculty. The interviews addressed topics such as daily work, sources of work satisfaction and success, work goals, activities, needs, and difficulties. The results provide an insight into the busy lives of STEM academics and can be useful to other efforts that aim to design for this user group. We discuss how we used these results, presented in the form of a persona, to evaluate existing design decisions and to create new features that would serve this audience.

Network for Computational Nanotechnology - a strategic plan for global knowledge transfer in research and education

The Network for Computational Nanotechnology (NCN) manages the science gateway nanoHUB.org, recognized as the worlds largest nanotechnology user facility, with over 2800 research and teaching resources in use by over 180,000 users annually. Resources consist of 220 simulation tools and nearly 2600 other content items ranging from podcasts of lectures to first time user guides for simulation tools to complete sets of university course materials. Simulation tools developed for research have been found to be used in the classroom and simple classroom tools are being used by researchers. With a global community spread across 172 countries, nanoHUB.org facilitates fast knowledge transfer across countries, disciplines, and communities. NCN follows a carefully planned strategy to lower barriers to this knowledge transfer and the growth and success of the site validates this strategy.

SensorWorld: Unified Touch-Based Access to Sensors Worldwide

Sensors play a critical role in monitoring environmental metrics, predicting natural hazards and developing sustainable policies. However, policy makers, engineers, scientists, students, and other stakeholders in fields such as environmental and natural resources management very rarely have access to real-world large-scale sensor data. This problem is compounded significantly by diverse and non-standard software interfaces to sensors. Our goal is to document and make available data from a large variety of real-world sensors through an intuitive interface. Our project addresses this problem by implementing a middleware framework and a client on iPhone to facilitate access to sensor data. Currently, we support a total of 1136 sensors from a variety of sources and the dataset is being rapidly expanded.

Utilizing visualization and feature selection methods to identify important learning objectives in a course

There have been numerous efforts to increase students academic success. One data-driven approach is to highlight the important learning objectives in a course. In this paper, we used visualization and three feature selection methods to highlight the important learning objectives in a course. Identifying important learning objectives as well as the relation among the learning objectives have multiple educational advantages. First, it informs the instructors and students of the important topics in the course; without learning them properly students will not be successful. Second, it highlights any inconsistencies in defining the learning objective, how they are being assessed, and design of the course. Thus, this approach can be used as a course design diagnostic tool.