The big data revolution has empowered universities with rich and complex digital data on their learners. Through digitally mediated activities conducted online or on-campus,
students leave behind digital footprints that can be mined to promote scholarly inquiry about student learning. My research contributes to personalising education and
translating traditional on-campus learning to authentic flexible learning in vibrant digital environments that better suit the needs and expectations of a digitally minded
generation. In particular, my aim is to employ exemplary techniques from the fields of machine learning, data mining, learning analytics, gamification, visualisation, education
and statistics to design, implement, validate and deliver solutions that use the digital footprints of learners towards improving students’ learning and enhancing their learning
experience. Some of the current projects that I
Recommendation in Personalised Peer-Learning EnvironmentsInprogress UQ Started in 2016
In recent years, personalisation in education has received significant attention. Personalisation is defined as tailoring a service or a product to accommodate a specific individual. An example of personalisation in a restaurant setting is having a chef make a dish specifically to the taste of an individual. An example of personalisation in education is creating learning objects (e.g., exercises or videos) specifically for the needs of an individual. Personalisation in education is still incoherent, expensive, and inchoate. On the other hand, recommendation is defined as making suggestions to accommodate a specific individual in making decisions. An example of recommendation in a restaurant setting is having a server suggest a dish from the menu to the taste of an individual. An example of recommendation in education is suggesting learning objects (e.g., exercises or videos) specifically for the needs of an individual. In this project, we propose to develop a platform that utilises recommender systems at its core to improve learning, enhance the learning experience of students and to move us a step closer towards truly personalising education. The recommender system will be designed to be extensible, broadly applicable, and scalable - supporting the diverse needs of the digitally-minded students of the 21st century.
The aim of the project is to design, implement, validate and deliver a course-level platform called RIPPLE (Recommendations in Personalised Peer-Learning Environments). RiPPLE maintains a repository of tagged multiple-choice questions (MCQs) that students can use for assessing and enhancing their learning. RiPPLE mines data collected by the system and employs exemplary techniques from the fields of machine learning, learning analytics and Recommender Systems for Technology Enhanced Learning (RecSysTEL) to:
- visualise students’ engagement, performance, and competencies
- recommend personalised questions that assist them in overcoming their knowledge gaps
A work-in-progress prototype of the system for use on desktops can be viewed at https://khosravi.uqcloud.net/ripple.
Reciprocal Peer Recommendation for Learning PurposesInprogress UQ Started in 2017
Using MOOC-like methodologies for Enhancing On-campus EducationInprogress UQ Started in 2017
Online educational platforms, and in particular, Massive Open Online Courses (MOOCs) have contributed significantly towards the spread and availability of high-quality, free educational for a huge domain of audience worldwide. UQx is a charter member of edX, a not-for-profit provider of MOOCs and delivers MOOCs on behalf of The University of Queensland. UQx has partnered with academics from The University of Queensland, thus far delivering 30 MOOCs to over 1.3 million learners since 2014 in either paced or self-paced mode. UQx MOOCs are delivered via the openedX platform which collates learner clickstream, assessment and discussion forum data.
As more and more on campus courses at UQ are seeking to transition to a flipped classroom model thereby incorporating flexible access to online learning content and activities, it is becoming imperative to analyse the learner data collected by UQx and use a data-driven approach to inform the design of on campus courses. In particular, there are some techniques and methodologies (such as use of peer-reviews, auto graders, self-evaluation, self-paced education) that are broadly used in MOOCs, that are not widely lised in face-to-face courses.
In this project, we aim to:
- uncover learner behaviour patterns contributing to course success or dropout
- compare paced and self-paced offerings of a MOOC
- analyse self-regulated learning strategies
- evaluate the impact of learning design choices on learner behavior and collaborative course participation
- support the diverse learning needs of the digitally minded students of the 21st century
- investigate how MOOC-like methodologies such as the use of peer reviews, automatic assessments, self-evaluation, and self-paced learning may be better utilised in on-campus education.
The project deliverables include extensions to the open source UQx Dashboard project.
Analysing the Learning Pathways of StudentsInprogress UQ Started in 2017
Educators continue to face significant challenges in providing high quality, post-secondary instruction in large classes including motivating and engaging diverse populations (e.g., academic ability and backgrounds, generational expectations), and providing helpful feedback and guidance.
Recent advancements in educational technologies (learning management systems, online discussion forums, peer-learning tools) coupled with new methods of course delivery (e.g. blended, flipped, MOOCs) provide significant opportunities for universities to deliver challenging, high quality, yet engaging curriculum for students. In this project, we examine the variations and similarities of student’s approaches to learning
Coupling Visualisation and Recommendation in Student-Facing Learning PlatformsInprogress UQ Started in 2017
There have been significant contributions from the learning analytics community on creating visualizations within student-facing learning dashboards to enhance learners' understanding of interactions with learning environments. While learning dashboards have been well-received in the research community, most of the developed dashboards, to date, have had limited ability in providing actionable insight for improving learning. In a separate body of work, inspired by the success of recommender systems, researchers from the educational data mining community have utilised the digital traces left by learners towards providing recommendation of resources that will assist students in overcoming their shortcomings. Despite the success of recommender systems in many other domains, they have not been well-adopted in the context of higher education. This may be because recommender systems often do not provide rationale for their recommendations. While this may be ok for recommending movies, for instance, inviting students to blindly follow recommendations without justification goes against the core mission of many universities, which is to develop visionary and critical thinkers. As a potential solution for addressing the twin challenges of visualisation without recommendation and recommendation without justification, we are in the process of designing, implementing, and validating an open-source student-facing learning platform called RiPPLE that couple visualisation and recommendation.
Video Analytics: Impact of Providing Personalised Video Feedback for AssessmentInprogress UQ Started in 2017
Use of videos has become a crucial aspect of both on-line and on-campus education. In addition to their direct benefit in communicating facts, demonstrating procedures, or providing personalised feedback to assist in mastery learning, use of videos allows for recording of digital traces of learners' interaction and navigation on a video, via a media player, that provide insights into student learning.
The aim of this project is to use video analytic techniques to investigate the impact of students' use of a system called vMarks, which provides on-the-go video feedback for assessment in large courses.
Some of the past projects that I have worked on are listed below. My past research interests involve graphical models, machine learning, mobile computing, and security.
Learning and Mining Education Datacompleted UBC 2016
The first-year engineering students at UBC use Blackboard to access the preparatory material for a course covering an introduction to programming. The data features of most interest to this project are the video screencasts, practice worksheets, pre-laboratory assignments, practice midterm exams and solutions, and solutions to the students’ actual previously written midterms. The main goals of this research project were to use machine learning and data mining algorithms to:
uncover relationships and patterns in the dataset that will help evolve the course (and potentially other courses) or provide useful recommendations to future students taking the course. For example, an interesting pattern is that students that watch screencasts at midday as opposed to late at night perform better on exam questions, and a beneficial recommendation is reporting that there is a limitation to the effectiveness of multiple viewings of a screencast (e.g., watching a screencast for a fifth time would probably not be a good use of a student’s time).
evaluate how students’ engagement and commitment to studying outside lectures, throughout the semester, is correlated with their performance on quizzes and exams, and whether it is possible to use linear regression or probabilistic graphical models to obtain a model that can accurately predict the exam grade of students based on their behavior throughout the semester
Peer Learning and Assessment Outside the Classroomcompleted UBC 2016
Use of clickers is a great way to foster effective engagement and peer learning among students; however, creating a meaningful and engaging environment for peer learning during lecture is time consuming, so unfortunately most of the time, the use of clickers needs to be shortened to provide time for the course content to be covered. As such, there is much interest and desire in exploring methods that allow students to have a similar peer learning experience outside the classroom.
PeerWise is a free web-based system in which students create multiple-choice questions, and answer, rate, and discuss questions created by their peers. It empowers students to generate their own multiple-choice questions and allows them to share them with one another.One of the shortcomings of PeerWise is that the created content cannot be exported and be used outside of the PeerWise platform. We have developed an open source framework that allows the content to be exported into a locally stored database and to be used without relying on PeerWise. The framework is easy to deploy and requires no intervention from the IT department. A prototype of the framework is available at https://github.com/hkhosrav/peerwise-flashcard
User Modeling on Smart Devicescompleted FusionPipe 2014-2015
Decisions on how to proceed when a device with confidential corporate information has been compromised are a function of the sensitivity of the stored data as well as policies of the host company. A traditional method of dealing with compromised devices aims to wipe the device remotely; however, such methods are particularly impractical when the signal is prevented. The two main components of this research project were the following:
Learning phase: We used machine learning techniques in data acquisition to extract features pertain- ing to a user’s interaction with a device to differentiate users’ patterns of interaction. Such data were utilized to build a probabilistic graphical model, which provides the basis for intelligent detection of unauthorized usage of the device. Sources such as meeting schedules, geographical data from GPS, available WiFis, internet traffic, patterns of how an application is used, and communication of the device with other devices such as smartwatches or laptops may be used for learning the model.
Compromise detection through inference: The model periodically updates its assessment from the state of the device. Once the model reaches a certain likelihood that the device is compromised, it triggers adaptive intervention, which allows the device to intelligently make decisions on its own. This includes remote/self termination of corporate information while fooling the attacker by revealing some false attractive information (honeypot methods). In case of availability of a signal the device may first transfer the prioritized information to a secure data store through cloud services, and then inform a reliable source, through email, of its current state and location.
We successfully designed a health agent, as a framework, that monitors and analyzes the behavior of a smart device and makes decisions on the health status of that device.
Behavioral Authentication on Smart Mobile Devices Using Probabilistic Graphical Modelscompleted UBC 2015-2016
This work is an extension over the “User Modeling on Smart Devices” project, in which we developed an algorithm that uses behavioral data such as touchscreen gestures, body movement styles, and holding patterns for authentication on mobile devices. We demonstrate that considering universality, permanence, performance, uniqueness, collectability, and circumvention behavioral biometrics, like physical biometrics, are ideal candidates for authentication. In addition, behavioral biometrics, unlike physical ones, have the advantage of supporting continuous authentication, being non-obtrusive, and cost effective. We developed a framework that uses probabilistic graphical models to learn the behavior of a user during an enrollment phase and is able to later classify whether the person interacting with the device is the true user or not. Experimental results from three datasets demonstrate that behavioral data from mobile devices are distinctive enough to serve as a biometric in verifying users.
Decoupling User Authentication and Data Encryption on Mobile Devicescompleted FusionPipe 2014-2015
Various forms of confidential data are stored on smart devices for the benefit of being accessible “on-the- go”, which creates the necessity for protecting “data-at-rest” on these devices. Data encryption with a randomly generated encryption key can be applied; however, since the device has to be able to work in off-line scenarios, the encryption key has to be stored on the device, along with data. In order to overcome this limitation, all major mobile platforms encrypt the encryption key with a so called key encryption key. The key encryption key is derived from a secret that is used to authenticate smartphone users (e.g., a PIN-code, or password). Unfortunately, PIN-codes are weak and are prone to brute-force attacks. In this work we presented and evaluated the KeyVault system that breaks the dependency between the security of data encryption and the secrecy of human memorable authentication secret. The main objective of the KeyVault system is to improve security of the data that are physically stored on the mobile devices. In order to achieve this we developed a framework to store the encryption keys on a separate device, and use wireless technologies to make them available when they are needed. This research led to the filing and approval of a full non-provisional patent titled ”Method and system for decoupling user authentication and data encryption on mobile devices”.
Conspiracy of Multiple Devices to form the Appearance of a Single Authenticated End Pointcompleted UBC 2015-2016
Using Cloud Computing for Disaster Recoverycompleted FusionPipe 2012-2013
Disaster recovery is an essential part of many businesses. A typical disaster recovery service works by replicating the data and the applications among multiple data centers. There are numerous technical challenges in traditional disaster recovery that make it expensive, slow, and complex for small or mid-sized enterprises. Traditional disaster recovery often entails very high cost, or provides very weak guarantees about the amount of data lost or time required to restart the operation after a failure. These challenges include the following: (1) successful recovery requires identical hardware configuration at the primary site and recovery sites, which often leads to expensive new server purchases, (2) complex system imaging tools and restore processes require special skills and resources which are costly and time consuming to integrate, and (3) snapshot based approaches limit the ability to retrieve individual files (or incremental changes) in the VM images.
The primary goal of the project was to develop disaster recovery for virtual machines as a cloud service. The idea is to use cloud computing instead of physical recovery sites for replication of applications and data. Our implementation allows two layers of recovery. The most delicate data and applications, tier 1 information, are stored on a hot standby recovery system, which is expected to start running within minutes after the disaster. Our implementation used a private cloud, operating on the open source OpenStack system, as the hot standby. Less delicate data and applications, tier 2 information, are stored on a warm standby recovery system. A warm standby recovery is expected to launch within a few hours of the disaster. We used Amazon Web Services as the warm standby.
Optimization of Cloud Usage on Amazon AWScompleted FusionPipe 2012-2013
The main goal of this project was to assist cloud consumers in optimizing cloud resource utilization. In particular, optimizing the costs incurred by provisioning infrastructure from the public cloud and the cost of software licenses, both on demand and perpetual, was targeted as the main aim of this project. This aim could be achieved by empowering the users of the cloud to gain better insight into the current cloud resources being provisioned and providing them with alternative solutions which could carry out the needs of the users with the same or even higher efficiency while imposing less cost in their usage bills. We successfully implemented an engine that optimizes the infrastructure and license utilization of the cloud users based on monitoring the cloud utilization on a real-time basis and using the knowledge base that we implemented.
Modelling Relational Statistics With Bayes Netscompleted SFU 2017-2012
My PhD dissertation was focused on graphical models in machine learning; in particular, on combining the scalability and efficiency of learning directed relational models, and the inference power and theoretical foundations of undirected relational models. The five years I spent on obtaining my PhD have been invaluable to my life and career, allowing me to acquire a solid foundation in machine learning, developing an appreciation for inquiry, innovation and research-based activities, and gaining experience in performing and reporting scientific research. Four of my significant contributions to research and development during my PhD was published in the Machine Learning Journal, which is one of the most prestigious journals in the field of machine learning. Executable, scripts, datasets, examples can be accessed at http://www.cs.sfu.ca/~oschulte/BayesBase/BayesBase.html