Multiple instance learning for classifying students in learning management systems

In this paper, a new approach based on multiple instance learning is proposed to predict student’s performance and to improve the obtained results using a classical single instance learning. Multiple instance learning provides a more suitable and optimized representation that is adapted to available information of each student and course eliminating the missing values that make difficult to find efficient solutions when traditional supervised learning is used. To check the efficiency of the new proposed representation, the most popular techniques of traditional supervised learning based on single instances are compared to those based on multiple instance learning. Computational experiments show that when the problem is regarded as a multiple instance one, performance is significantly better and the weaknesses of single-instance representation are overcome.     

Stable Label-Specific Features Generation for Multi-Label Learning via Mixture-Based Clustering Ensemble

Multi-label learning deals with objects associated with multiple class labels, and aims to induce a predictive model which can assign a set of relevant class labels for an unseen instance. Since each class might possess its own characteristics, the strategy of extracting label-specific features has been widely employed to improve the discrimination process in multi-label learning, where the predictive model is induced based on tailored features specific to each class label instead of the identical instance representations. As a representative approach, LIFT generates label-specific features by conducting clustering analysis. However, its performance may be degraded due to the inherent instability of the single clustering algorithm. To improve this, a novel multi-label learning approach named SENCE (stable label-Specific features gENeration for multi-label learning via mixture-based Clustering Ensemble) is proposed, which stabilizes the generation process of label-specific features via clustering ensemble techniques. Specifically, more stable clustering results are obtained by firstly augmenting the original instance repre-sentation with cluster assignments from base clusters and then fitting a mixture model via the expectation-maximization (EM) algorithm. Extensive experiments on eighteen benchmark data sets show that SENCE performs better than LIFT and other well-established multi-label learning algorithms.      

Adding value to student learning: Integrating the hybrid library into the virtual learning environment

A twelve‐month project at Leeds University Library explored the integration of library and information resources into the undergraduate learning environment. Five modules (585 students) were included in the trial. Working closely with academic course leaders, library staff created for each module a tailored environment which would support student learning by providing appropriate information resources within the virtual learning environment. Each ‘hybrid library ‘ focused on core materials such as reading lists and recommended readings, but also acted as a gateway to wider information resources such as the OPAC and relevant online databases and websites.

Evaluation was carried out using student questionnaires, focus groups and semi‐structured interviews with academic staff. Analysis of the questionnaires revealed that 79% of respondents had used the facility, and the vast majority had found it relevant and useful. 27% had accessed it from off campus. Electronic reading lists and digitised core readings were particularly valued, but 32% had used the links to external databases and web‐sites. The response from academics was equally positive, with a perceived improvement in student learning. Encouragement by academic staff and the juxtaposition of generic networked information sources alongside core course materials were key factors in encouraging a more outward and exploratory approach in students.     

Comparing strategies for modeling students learning styles through reinforcement learning in adaptive and intelligent educational systems: An experimental analysis

A huge number of studies attest that learning is facilitated if teaching strategies are in accordance with students learning styles, making the learning process more effective and improving students performances. In this context, this paper presents an automatic, dynamic and probabilistic approach for modeling students learning styles based on reinforcement learning. Three different strategies for updating the student model are proposed and tested through experiments. The results obtained are analyzed, indicating the most effective strategy. Experiments have shown that our approach is able to automatically detect and precisely adjust students’ learning styles, based on the non-deterministic and non-stationary aspects of learning styles. Because of the probabilistic and dynamic aspects enclosed in automatic detection of learning styles, our approach gradually and constantly adjusts the student model, taking into account students’ performances, obtaining a fine-tuned student model.     

Content-free collaborative learning modeling using data mining

Modeling user behavior (user modeling) via data mining faces a critical unresolved issue: how to build a collaboration model based on frequent analysis of students in order to ascertain whether collaboration has taken place. Numerous human-based and knowledge-based solutions to this problem have been proposed, but they are time-consuming or domain-dependent. The diversity of these solutions and their lack of common characteristics are an indication of how unresolved this issue remains. Bearing this in mind, our research has made progress on several fronts. First, we have found supportive evidence, based on a collaborative learning experience with hundreds of students over three consecutive years, that an approach using domain independent learning that is transferable to current e-learning platforms helps both students and teachers to manage student collaboration better. Second, the approach draws on a domain-independent modeling method of collaborative learning based on data mining that helps clarify which user-modeling issues are to be considered. We propose two data mining methods that were found to be useful for evaluating student collaboration, and discuss their respective advantages and disadvantages. Three data sources to generate and evaluate the collaboration model were identified. Third, the features being modeled were made accessible to students in several meta-cognitive tools. Their usage of these tools showed that the best approach to encourage student collaboration is to show only the most relevant inferred information, simply displayed. Moreover, these tools also provide teachers with valuable modeling information to improve their management of the collaboration. Fourth, an ontology, domain independent features and a process that can be applied to current e-learning platforms make the approach transferable and reusable. Fifth, several open research issues of particular interest were identified. We intend to address these open issues through research in the near future.     

A NEW E‐LEARNING ACHIEVEMENT EVALUATION MODEL BASED ON ROUGH SET AND SIMILARITY FILTER

The trend of utilizing information and Internet technologies as teaching and learning tools is rapidly expanding into education. E‐learning is one of the most popular learning environments in the information era. The Internet enables students to learn without limitations of space and time. Furthermore, the learners can repeatedly review the context of a course without the barrier of distance. Recently, student‐centered instruction has become the primary trend in education, and the e‐learning system, which is considered with regard to of personalization and adaptability, is more and more popular. By means of e‐learning systems, teachers can adjust the learning schedule instantly for each learner according to a student's achievements and build more adaptive learning environments. Sometimes, teachers give biased assessments of students’ achievements under uncontrollable conditions (i.e., tiredness, preference) and are in dire need of overcoming this predicament. To solve the drawback mentioned, a new model to evaluate learning achievements based on rough set and similarity filter is proposed. The proposed model includes four facets: (1) select important features (attributes) to enhance classification performance by feature selection methods; (2) utilize minimal entropy principle approach (MEPA) to fuzzify the quantitative data; (3) select linguistic values for each feature and delete inconsistent data using the similarity threshold (similarity filter); and (4) generate rules based on rough set theory (RST). The practical e‐learning achievement data sets are collected by an e‐learning online examination system from a university in Taiwan. To verify our model, the performances of the proposed model are compared with the listing models. Results of this study demonstrate that the proposed model outperforms the listing models.     

A divide-and-conquer recursive approach for scaling up instance selection algorithms

Instance selection is becoming more and more relevant due to the huge amount of data that is being constantly produced. However, although current algorithms are useful for fairly large datasets, scaling problems are found when the number of instances is of hundreds of thousands or millions. In the best case, these algorithms are of efficiency O(n ²), n being the number of instances. When we face huge problems, scalability is an issue, and most algorithms are not applicable. This paper presents a divide-and-conquer recursive approach to the problem of instance selection for instance based learning for very large problems. Our method divides the original training set into small subsets where the instance selection algorithm is applied. Then the selected instances are rejoined in a new training set and the same procedure, partitioning and application of an instance selection algorithm, is repeated. In this way, our approach is based on the philosophy of divide-and-conquer applied in a recursive manner. The proposed method is able to match, and even improve, for the case of storage reduction, the results of well-known standard algorithms with a very significant reduction of execution time. An extensive comparison in 30 datasets form the UCI Machine Learning Repository shows the usefulness of our method. Additionally, the method is applied to 5 huge datasets with from 300,000 to more than a million instances, with very good results and fast execution time.     

Activity index model for self-regulated learning with learning analysis in a TEL environment

Various learner-oriented teaching–learning models are spreading along with development of the technology-enhanced learning (TEL) environment and the spread of the massive open online course (MOOC). Vast amounts of various data are being created and accumulated from learning activities based on the TEL environment. Also, a self-regulated learning ability is required in the MOOC environment because the learning process is constituted on students making decisions by themselves. Accordingly, this study is aimed at suggesting an activity index model based on self-regulated learning and an activity index based on self-regulated learning. It is intended to provide a means to collect proof of what influences the teaching–learning activity. This model is intended to set a learning activity standard on the basis of general activity, interaction activity, and achievement activity by students. It will be possible to analyze the student’s participation level based on the activity index, which is based on self-regulated learning, to induce participation in the teaching–learning activity, and to recommend more appropriate learning activity elements. The student data are divided into score-related, time-related, and count-related groups for applications. The stabilization of the data was confirmed through time series analysis. In multiple regression analysis, the academic achievement element was set by the target variable, and the relationships among explanatory variables were confirmed. It was understood from the explanatory variables that similar student groups were highly concerned with notice participation in the learning activity. It will be possible to analyze the students’ participation levels, induce participation in the teaching–learning activities, and recommend more appropriate learning activity elements on the basis of an activity index based on self-regulated learning.

     

Designing a Library Information Literacy Program Using Threshold Concepts,Student Learning Theory,and Millennial Research in the Development of Information Literacy Sessions

This article will discuss Ferrum College's approach toward developing and delivering varied one-shot information literacy instruction sessions based on knowledge of millennial students’ information retrieval practices and on literature regarding threshold concepts and teaching and learning theories. Identifying threshold information literacy concepts helped the library to develop session specific learning outcomes to meet the needs of their students. Moreover, knowledge of learning theory and millennial learning preferences aided in developing lesson plans and activities that encouraged student mastery of these concepts. Ultimately, this approach contributed to a more comprehensive information literacy program that could be scaffolded over multiple instruction visits.     

Adaptive load balancing in learning-based approaches for many-core embedded systems

Adaptive routing algorithms improve network performance by distributing traffic over the whole network. However, they require congestion information to facilitate load balancing. To provide local and global congestion information, we propose a learning method based on dual reinforcement learning approach. This information can be dynamically updated according to the changing traffic condition in the network by propagating data and learning packets. We utilize a congestion detection method which updates the learning rate according to the congestion level. This method calculates the average number of free buffer slots in each switch at specific time intervals and compares it with maximum and minimum values. Based on the comparison result, the learning rate sets to a value between 0 and 1. If a switch gets congested, the learning rate is set to a high value, meaning that the global information is more important than local. In contrast, local is more emphasized than global information in non-congested switches. Results show that the proposed approach achieves a significant performance improvement over the traditional Q-routing, DRQ-routing, DBAR and Dynamic XY algorithms.     

Instance selection for regression by discretization

An important step in building expert and intelligent systems is to obtain the knowledge that they will use. This knowledge can be obtained from experts or, nowadays more often, from machine learning processes applied to large volumes of data. However, for some of these learning processes, if the volume of data is large, the knowledge extraction phase is very slow (or even impossible). Moreover, often the origin of the data sets used for learning are measure processes in which the collected data can contain errors, so the presence of noise in the data is inevitable. It is in such environments where an initial step of noise filtering and reduction of data set size plays a fundamental role. For both tasks, instance selection emerges as a possible solution that has proved to be useful in various fields. In this paper we focus mainly on instance selection for noise removal. In addition, in contrast to most of the existing methods, which applied instance selection to classification tasks (discrete prediction), the proposed approach is used to obtain instance selection methods for regression tasks (prediction of continuous values). The different nature of the value to predict poses an extra difficulty that explains the low number of articles on the subject of instance selection for regression.More specifically the idea used in this article to adapt to regression problems “classic” instance-selection algorithms for classification is as simple as the discretization of the numerical output variable. In the experimentation, the proposed method is compared with much more sophisticated methods, specifically designed for regression, and shows to be very competitive.The main contributions of the paper include: (i) a simple way to adapt to regression instance selection algorithms for classification, (ii) the use of this approach to adapt a popular noise filter called ENN (edited nearest neighbor), and (iii) the comparison of this noise filter against two other specifically designed for regression, showing to be very competitive despite its simplicity.     

Development of a diagnostic system using a testing-based approach for strengthening student prior knowledge

Students learn new instructions well by building on relevant prior knowledge, as it affects how instructors and students interact with the learning materials. Moreover, studies have found that good prior knowledge can enable students to attain better learning motivation, comprehension, and performance. This suggests it is important to assist students in obtaining the relevant prior knowledge, as this can enable them to engage meaningfully with the learning materials. Tests are often used to help instructors assess students’ prior knowledge. Nevertheless, conventional testing approaches usually assign only a score to each student, and this may mean that students are unable to realize their own individual weaknesses. To address this problem, instructors can diagnose the test results to provide more detailed information to each student, but this is obviously a time-consuming process. Therefore, this study proposes a testing-based diagnosis system to assist instructors and students in diagnosing and strengthening prior knowledge before new instruction is undertaken. Furthermore, an experiment was conducted to evaluate the effectiveness of the proposed approach in an interdisciplinary course, since several studies have indicated that students learn more and better in such courses when applying relevant prior knowledge to what they are learning. The experimental results show that the developed system is able to effectively diagnose students’ prior knowledge and enhance their learning motivation and performance on an interdisciplinary course. In addition, two diagnostic evaluations were also conducted to assess whether the diagnoses given by the system were consistent with the decisions of experts. The results demonstrate that the proposed system can effectively assist instructors and students in diagnosing and strengthening prior knowledge before new instruction is undertaken, since the diagnoses produced by the system were broadly consistent with those of experts.     

Mining LMS data to develop an “early warning system” for educators: A proof of concept

Earlier studies have suggested that higher education institutions could harness the predictive power of Learning Management System (LMS) data to develop reporting tools that identify at-risk students and allow for more timely pedagogical interventions. This paper confirms and extends this proposition by providing data from an international research project investigating which student online activities accurately predict academic achievement. Analysis of LMS tracking data from a Blackboard Vista-supported course identified 15 variables demonstrating a significant simple correlation with student final grade. Regression modelling generated a best-fit predictive model for this course which incorporates key variables such as total number of discussion messages posted, total number of mail messages sent, and total number of assessments completed and which explains more than 30% of the variation in student final grade. Logistic modelling demonstrated the predictive power of this model, which correctly identified 81% of students who achieved a failing grade. Moreover, network analysis of course discussion forums afforded insight into the development of the student learning community by identifying disconnected students, patterns of student-to-student communication, and instructor positioning within the network. This study affirms that pedagogically meaningful information can be extracted from LMS-generated student tracking data, and discusses how these findings are informing the development of a customizable dashboard-like reporting tool for educators that will extract and visualize real-time data on student engagement and likelihood of success.     

一种聚类型基于示例学习新方法

借鉴聚类思想和万有引力计算方法,提出了解决基于示例学习中两个关键问题的新思路,这两个新思路分别是,利用示例邻近同类其它示例数目来描述该示例潜在预测能力,以及利用实例质量来帮助更加准确地预测新实例类别。据此构造了一种聚类型基于示例学习的新方法,并利用标准机器学习数据库中3个复杂数据样本,对所提方法的性能进行实验检测,有关的对比实验结果表明,所提方法在实例预测能力以及学习结果占用空间有效性方面,均优越其它多种基于示范学习方法。     

Two learning approaches for a rule-based intuitive reasoner

This paper is a discussion of two continuous learning approaches for improving classification accuracy for an intuitive reasoner algorithm. The reasoner predicted the value of a given target variable by multiple iterations of forward-chained, rule-based inference. Each rule in the reasoner’s rule set had associated with it a weight, referred to here as “Strength of Belief” (SB). The value of SB of a rule indicated the certainty level of that rule. In each iteration of reasoning, any instances of similar values for a given variable were replaced by a single consolidated datum and the SB associated with the consolidated datum was increased. At the end of the reasoning process, the class (value) of the target variable which had the highest SB was reported as the conclusion. The rule set for the reasoner was generated based on a training data set that contained 80% of the data in a weather database comprising 50 years worth of hourly measurements for 54 weather variables. Each rule was induced based on only a small subset of the weather data. The intuitive reasoner was tested by using the induced rules to predict a number of pre-selected target variables using 275 test cases created from the test data. The first continuous learning approach was to identify relevant input variables for the reasoner, and the second was to rebalance the rule set used by the reasoner by adjusting the SB associated with each of the rules. Because of the way the rules were induced, the resulting rules did not contain any information about the relevance of the 53 possible input variables to the task of predicting a given target variable for previously unseen cases. A method was developed to identify which input variables were most relevant to the task based on the induced rule set. This method resulted in higher prediction accuracy of the intuitive reasoner than using a set of randomly chosen input variables for four of six target variables. The second continuous learning approach was intended to address the class imbalance problem in the rule set. The intuitive reasoner appeared to over-fit classes (values) which had frequent representation in the rule set. To address this problem, a heuristic was developed that generated adjustment factors for the SB values of the rules. The use of this heuristic improved the classification accuracy of the intuitive reasoner for four of the six target variables.     

Multi-view Instance Attention Fusion Network for classification

Multi-view learning for classification has achieved a remarkable performance compared with the single-view based methods. Inspired by the instance based learning which directly regards the instance as the prior and well preserves the valuable information in different instances, a Multi-view Instance Attention Fusion Network (MvIAFN) is proposed to efficiently exploit the correlation across both instances and views. Specifically, a small number of instances from different views are first sampled as the set of templates. Given an additional instance and based on the similarities between it and the selected templates, it can be re-presented by following an attention strategy. Thanks for this strategy, the given instance is capable of preserving the additional information from the selected instances, achieving the purpose of extracting the instance-correlation. Additionally, for each sample, we not only perform the instance attention in each single view but also get the attention across multiple views, allowing us to further fuse them to obtain the fused attention for each view. Experimental results on datasets substantiate the effectiveness of our proposed method compared with state-of-the-arts.     

Scaffolding student online discussions using past discussions: PedaBot studies

We are developing instructional tools that will help students and instructors use discussion boards more effectively, with an emphasis on automatically assessing discussion activities and promoting student discussion participation and learning. In this paper, we present a discussion scaffolding tool that exploits natural language processing and information retrieval techniques. The PedaBot tool is designed to aid student knowledge acquisition, promote reflection about course topics and encourage student participation in discussions. It dynamically processes student discussions and presents related discussions and document from a knowledge base of past discussions and course materials. This paper describes the system and presents a comparative analysis of the information retrieval techniques used to respond to free-form student discussions, including a combination of topic profiling, term frequency-inverse document frequency, and latent semantic analysis. Responses are presented as annotated links that students can follow and rate for usefulness. The tool has been integrated into a live discussion board and has been used by an undergraduate computer science course for three semesters. We report current studies of PedaBot from its usages based on student viewings, student ratings, and a small survey. Initial results indicate that there is a high level of student interest in the feature and that its responses are moderately relevant to student discussions. We are exploring more opportunities to exposing the tool to students.     

聚合邻域信息的联合知识表示模型

知识表示学习在关系抽取、自动问答等自然语言处理任务中获得了广泛关注,该技术旨在将知识库中的实体与关系表示为稠密低维实值向量.然而,已有的模型在建模知识库中的三元组时,或是忽略三元组的邻域信息,导致无法处理关联知识较少的罕见实体,或是在引入邻域信息时不能自适应地为每个实体抽取最相关的邻节点属性,导致引入了冗余信息.基于以...     

Reality is our laboratory: communities of practice in applied computer science

The present paper presents a longitudinal study of the course 'High-tech Entrepreneurship and New Media'. The course design is based on socio-cultural theories of learning and considers the role of social capital in entrepreneurial networks. By integrating student teams into the communities of practice of local start-ups, we offer learning opportunities to students, companies and academia. The student teams are connected to each other and to their supervisors in academia and practice through a community-system. Moreover, the course is accompanied by a series of lectures and group discussions. In this paper we want to present our experiences and to reflect upon the design changes between the first and the second instance of the course. The evaluation of the course showed that the work on real-world problems and the collaboration in teams together with partners from start-up companies were evaluated as very positive, although design flaws, and cultural and professional diversities limited the success of the first instance in 2001. For the second course in 2002, the didactical design was improved significantly according to evaluation results, which brought evidence that the design changes resulted in better collaborative practices and more stable relationships between start-up companies and students. Furthermore, it was found that especially the differences in cultural background and different historical experiences between the two distinct groups of 'students' and 'entrepreneurs' might make processes of social identification more difficult and, therefore, successful community-building less likely'.     

Incremental Multi-Label Learning with Active Queries

In multi-label learning,it is rather expensive to label instances since they are simultaneously associated with multiple labels.Therefore,active learning,which reduces the labeling cost by actively querying the labels of the most valuable data,becomes particularly important for multi-label learning.A good multi-label active learning algorithm usually consists of two crucial elements:a reasonable criterion to evaluate the gain of querying the label for an instance,and an effective classification model,based on whose prediction the criterion can be accurately computed.In this paper,we first introduce an effective multi-label classification model by combining label ranking with threshold learning,which is incrementally trained to avoid retraining from scratch after every query.Based on this model,we then propose to exploit both uncertainty and diversity in the instance space as well as the label space,and actively query the instance-label pairs which can improve the classification model most.Extensive experiments on 20 datasets demonstrate the superiority of the proposed approach to state-of-the-art methods.