Session Clustering Research Articles

A Decision-oriented Approach to Evaluating a Leadership Curriculum in Fellowship.

Graduate Medical Education (GME) emphasizes the role of the annual program evaluation to identify opportunities, create action plans, and track improvements longitudinally. There is a lack of a systematic approach to the evaluation of educational curricula. Comprehensive curriculum evaluation can inform the educators about specific modifications to achieve high standards, desired outcomes, and the anticipated objectives. To evaluate a leadership in quality improvement program in a pulmonary/critical care fellowship training program using the context, input, process, product (CIPP) model. The CIPP model, given its focus on evaluating different aspects of a program, provides concrete and targeted feedback to guide improvement decisions. Evaluation questions addressing the four focused areas were created, pilot tested, and revised. The questions were framed toward optimization of alignment (e.g., program activities with stated objectives, program goals with theoretical perspective, program curriculum with trainee needs) and gaining information about the efficacy of the program in achieving the desired outcomes. To enhance the validity of the results, we triangulated the data-gathering approach by administering surveys and conducting interviews and focus groups by random selection from the eligible participants. Qualitative data were transcribed, coded, and categorized into themes aligning with the four aspects of the CIPP model. We interviewed 9 participants and conducted three focus groups with 20 participants. The surveys provided vital quantitative information that was cross-verified with the qualitative data; 23 of the 25 (92%) participants completed the survey. The results of qualitative thematic analysis were organized in the CIPP format. The context evaluation of the program revealed that the fellows and faculty were unfamiliar with the guiding principles of the course. The input evaluation highlighted the competing interests that hampered the engagement of the fellows during the evening weekly report-outs. The process evaluation revealed clustering of didactic sessions at the start of the course. The product evaluation stressed the difficulty in completing the quality improvement projects in the allotted timeframe. Conducting a robust evaluation of an educational curriculum provides insights into gaps in the various stages of the program. Time and resources needed for conducting evaluation by using the CIPP model should be considered.

Measuring self‐regulated learning in a junior high school mathematics classroom: Combining aptitude and event measures in digital learning materials

AbstractBackgroundMeasurement of students' self‐regulation skills is an active topic in education research, as effective assessment helps devising support interventions to foster academic achievement. Measures based on event tracing usually require large amounts of data (e.g., MOOCs and large courses), while aptitude measures are often qualitative and need careful interpretation. Precise and interpretable evaluation of self‐regulation skills in a normal K‐12 classroom thus poses a challenge.ObjectivesThe present study proposes and explores a learning analytics method of combining aptitude and event measures to evaluate student's self‐regulation skills.MethodsAn explorative learning analytics study was conducted in a junior high school mathematics class ( students), using a three‐lesson intervention with digital learning materials. Students first assessed their self‐regulation skills with a self‐report questionnaire, after which trace logs and observations of student behaviour were collected. Learning sessions were extracted from trace logs, clustered, and linked to learning strategies. Students were clustered by the self‐report results and learning behaviour profiles. Session clusters, student behaviour clusters and assignment grades were also tested for association.Results and ConclusionsThe detected session and student behaviour types were linked to learning tactics and strategies found in prior studies. Additionally, association was found between self‐reported self‐regulation skills and the student behaviour obtained from trace logs.ImplicationsThe results demonstrate the feasibility of concurrently using aptitude and event measures on a classroom scale, providing teachers with a tool to evaluate and support self‐regulated learning. Combined with further measures like predictive learning analytics, teachers can obtain an early and highly interpretable picture of at‐risk students in their classes.

ProGraph: Robust Network Traffic Identification With Graph Propagation

Network traffic identification is critical for effective network management. Existing methods mostly focus on invariant network environments with stable attribute distributions. Unfortunately, however, they can hardly be adaptive to the variation of practical networks and suffer from significant performance degradation. This problem largely stems from the over-dependence of existing methods on the vulnerable side-channel features. To address this issue, in this paper we propose a graph-based approach, namely ProGraph, to ensure robust network traffic classification among various network environments. The core idea of ProGraph is to construct a correlation graph with session clusters aggregated from different networks, based on which graph propagation can be effectively implemented to predict labels of testing nodes in an iterative manner. ProGraph enhances the correlation between clusters of the same class to provide reliable paths for label dissemination from the labeled clusters to the testing ones. It is encouraging to see that the proposed ProGraph achieves an accuracy of 92.25% in networks with constant attributes, while remaining stable with the accuracy of 90.89% when deployed in different networks, which significantly outperforms the state-of-the-art approaches. Meanwhile, ProGraph can accurately identify the novel classes which do not exist in the training dataset, with an AUC of 95.11. Last but not least, a carefully constructed dataset, namely CrossNet2021, containing network traffic of 20 classes of applications from two distinct networking scenarios, is made publicly available to support further research.

An Efficient Fuzzy Clustering Algorithm for Mining User Session Clusters on Web Log Data

Data mining is extremely vital to get important information from the web. Additionally, web usage mining (WUM) is essential for companies. WUM permits organizations to create rich information related to the eventual fate of their commercial capacity. The utilization of data that is assembled by Web Usage Mining gives the organizations the capacity to deliver results more compelling to their organizations and expanding of sales. Client access patterns can be mined from web access log information using Web Usage Mining (WUM) techniques. Because there are so many end-user sessions and URL resources, the size of web user session data is enormous. Human communications and non-deterministic browsing patterns increment equivocalness and dubiousness of client session information. The fuzzy set-based approach can solve most of the challenges listed above. This paper proposes an efficient Fuzzy Clustering algorithm for mining client session clusters from web access log information to find the groups of client profiles. In addition, the methodologies to preprocess the net log data as well as data cleanup client identification and session identification are going to be mentioned. This incorporates the strategy to do include choice (or dimensionality decrease) and meeting weight task assignments

Xatu

The performance of Adaptive Bitrate (ABR) algorithms for video streaming depends on accurately predicting the download time of video chunks. Existing prediction approaches (i) assume chunk download times are dominated by network throughput; and (ii) apriori cluster sessions (e.g., based on ISP and CDN) and only learn from sessions in the same cluster. We make three contributions. First, through analysis of data from real-world video streaming sessions, we show (i) apriori clustering prevents learning from related clusters; and (ii) factors such as the Time to First Byte (TTFB) are key components of chunk download times but not easily incorporated into existing prediction approaches. Second, we propose Xatu, a new prediction approach that jointly learns a neural network sequence model with an interpretable automatic session clustering method. Xatu learns clustering rules across all sessions it deems relevant, and models sequences with multiple chunk-dependent features (e.g., TTFB) rather than just throughput. Third, evaluations using the above datasets and emulation experiments show that Xatu significantly improves prediction accuracies by 23.8% relative to CS2P (a state-of-the-art predictor). We show Xatu provides substantial performance benefits when integrated with multiple ABR algorithms including MPC (a well studied ABR algorithm), and FuguABR (a recent algorithm using stochastic control) relative to their default predictors (CS2P and a fully connected neural network respectively). Further, Xatu combined with MPC outperforms Pensieve, an ABR based on deep reinforcement learning.

Semantic Based Weighted Web Session Clustering Using Adapted K-Means and Hierarchical Agglomerative Algorithms

The WWW has a big number of pages and URLs that supply the user with a great amount of content. In an intensifying epoch of information, analysing users browsing behaviour is a significant affair. Web usage mining techniques are applied to the web server log to analyse the user behaviour. Identification of user sessions is one of the key and demanding tasks in the pre-processing stage of web usage mining. This paper emphasizes on two important fallouts with the approaches used in the existing session identification methods such as Time based and Referrer based sessionization. The first is dealing with comparing of current request’s referrer field with the URL of previous request. The second is dealing with session creation, new sessions are created or comes in to one session due to threshold value of page stay time and session time. So, authors developed enhanced semantic distance based session identification algorithm that tackles above mentioned issues of traditional session identification methods. The enhanced semantic based method has an accuracy of 84 percent, which is higher than the Time based and Time-Referrer based session identification approaches. The authors also used adapted K-Means and Hierarchical Agglomerative clustering algorithms to improve the prediction of user browsing patterns. Clusters were found using a weighted dissimilarity matrix, which is calculated using two key parameters: page weight and session weight. The Dunn Index and Davies-Bouldin Index are then used to evaluate the clusters. Experimental results shows that more pure and accurate session clusters are formed when adapted clustering algorithms are applied on the weighted sessions rather than the session obtained from traditional sessionization algorithms. Accuracy of the semantic session cluster is higher compared with the cluster of sessions obtained using traditional sessionization.

Xatu: Richer Neural Network Based Prediction for Video Streaming

The performance of Adaptive Bitrate (ABR) algorithms for video streaming depends on accurately predicting the download time of video chunks. Existing prediction approaches (i) assume chunk download times are dominated by network throughput; and (ii) apriori cluster sessions (e.g., based on ISP and CDN) and only learn from sessions in the same cluster. We make three contributions. First, through analysis of data from real-world video streaming sessions, we show (i) apriori clustering prevents learning from related clusters; and (ii) factors such as the Time to First Byte (TTFB) are key components of chunk download times but not easily incorporated into existing prediction approaches. Second, we propose Xatu, a new prediction approach that jointly learns a neural network sequence model with an interpretable automatic session clustering method. Xatu learns clustering rules across all sessions it deems relevant, and models sequences with multiple chunk-dependent features (e.g., TTFB) rather than just throughput. Third, evaluations using the above datasets and emulation experiments show that Xatu significantly improves prediction accuracies by 23.8% relative to CS2P (a state-of-the-art predictor). We show Xatu provides substantial performance benefits when integrated with multiple ABR algorithms including MPC (a well studied ABR algorithm), and FuguABR (a recent algorithm using stochastic control) relative to their default predictors (CS2P and a fully connected neural network respectively). Further, Xatu combined with MPC outperforms Pensieve, an ABR based on deep reinforcement learning.

Identifying legitimate Web users and bots with different traffic profiles — an Information Bottleneck approach

Recent studies reported that about half of Web users nowadays are intelligent agents (Web bots). Many bots are impersonators operating at a very high sophistication level, trying to emulate navigational behaviors of legitimate users (humans). Moreover, bot technology continues to evolve which makes bot detection even harder. To deal with this problem, many advanced methods for differentiating bots from humans have been proposed, a large part of which relies on supervised machine learning techniques. In this paper, we propose a novel approach to identify various profiles of bots and humans which combines feature selection and unsupervised learning of HTTP-level traffic patterns to develop a user session classification model. Session clustering is performed with the agglomerative Information Bottleneck (aIB) algorithm, as well as with some other reference algorithms. The model is then used to classify new sessions to one of the profiles and to label the sessions as performed by bots or humans. An extensive experimental study, based on real server log data, demonstrates the ability of aIB clustering to distinguish user profiles and confirms high performance of the classification model in terms of accuracy, F1, recall, and precision.

A Computerized Text and Cluster Analysis Approach to Psychotherapy Talk

This paper illustrates an analytical approach combining LIWC, a computer text-analytic application, with cluster analysis techniques to explore ‘language styles’ in psychotherapy across sessions in time. It categorizes session transcripts into distinct clusters or styles based on linguistic (di)similarity and relates them to sessional progression, thus providing entry points for further qualitative exploration. In the first step, transcripts of four illustrative therapist-client dyads were scored under ten LIWC variables including ‘analytic thinking’, ‘clout’, ‘authenticity’, ‘emotional tone’, and pronoun types. In the next step, agglomerative hierarchical clustering uncovered distinct session clusters that are differently distributed in each dyad. The relationships between these clusters and the chronological progression of sessions were then further discussed in context as contrastive exemplars. Applications, limitations and future directions are highlighted.

A review of alignment based similarity measures for web usage mining

In order to understand web-based application user behavior, web usage mining applies unsupervised learning techniques to discover hidden patterns from web data that captures user browsing on web sites. For this purpose, web session clustering has been among the most popular approaches to group users with similar browsing patterns that reflect their common interest. An adequate web session clustering implementation significantly depends on the measure that is used to evaluate the similarity of sessions. An efficient approach to evaluate session similarity is sequence alignment, which is known as the task of determining the similarity of elements between sequences. In this paper, we review and compare sequence alignment-based measures for web sessions, and also discuss sequence similarity measures that are not alignment-based. This review also provides a perspective of sequence similarity measures that manipulate web sessions in usage clustering process.

An automatic test case generation method based on user session and agglutinate hierarchical clustering algorithm

An automatic test case generation method based on user session and agglutinate hierarchical clustering algorithm

A Novel Session Clustering Approach Based On User Feed Back

Most search engines (SE) adopt the keyword-based query method. The system would finally produce many complicated cross-field query outcomes when systems did not respectively manage to classify user latent query requisition and specifying fields. Web user session clustering is very important in web usage mining for web personalization. This paper proposes a novel session clustering approach based on user feedback. For every specific user, we get two semantic relationships between userquery and their search click information, such as expanded query contents and selected documents. In this way, user-session semantic similarity can be calculated using search click information, then user session can be clustered and disambiguated based on user's interests. Experiments have been conducted and the results have shown that our clustering approach is capable of clustering web user sessions with similar interests and topics.

A subtractive medoids selection based fuzzy relational clustering of augmented web user sessions

A subtractive medoids selection based fuzzy relational clustering of augmented web user sessions

Fuzzy c-Least Medians clustering for discovery of web access patterns from web user sessions data

Mining web usage data of e-business organizations is essential to provide knowledge about clients’ web utilization patterns, which can help these businesses in landing at vital business choices. Because of non-deterministic web access behavior of web clients, web user session data is usually noisy and imperfect. Such imperfection has a negative impact on pattern discovery process. One of the real issues associated with the prevalently used Fuzzy c-Means (FCM) and Fuzzy c-Medoids (FCMdd) methods is that they are not robust against the noise, because a single outlier object could lead to a very different clustering result. In this research we propose a robust Fuzzy c-Least Medians (FCLMdn) clustering framework to deal with the user session data contaminated with noise and outlier user session objects, with the objective of improving the quality of the extracted patterns. To deal with the high dimensionality of user session data which may contain noise and outliers, a fuzzy set theoretic approach for assigning fuzzy weights to user sessions and associated URLs has been proposed. Our results clearly indicate that quality of user session clusters formed using FCLMdn algorithm is much better than those using FCM and FCMdd algorithms in terms of various cluster validity indices.

Inpatient Hospital Factors and Resident Time With Patients and Families.

To define hospital factors associated with proportion of time spent by pediatric residents in direct patient care. We assessed 6222 hours of time-motion observations from a representative sample of 483 pediatric-resident physicians delivering inpatient care across 9 pediatric institutions. The primary outcome was percentage of direct patient care time (DPCT) during a single observation session (710 sessions). We used one-way analysis of variance to assess a significant difference in the mean percentage of DPCT between hospitals. We used the intraclass correlation coefficient analysis to determine within- versus between-hospital variations. We compared hospital characteristics of observation sessions with ≥12% DPCT to characteristics of sessions with <12% DPCT (12% is the DPCT in recent resident trainee time-motion studies). We conducted mixed-effects regression analysis to allow for clustering of sessions within hospitals and accounted for correlation of responses across hospital. Mean proportion of physician DPCT was 13.2% (SD = 8.6; range, 0.2%-49.5%). DPCT was significantly different between hospitals (P < .001). The intraclass correlation coefficient was 0.25, indicating more within-hospital than between-hospital variation. Observation sessions with ≥12% DPCT were more likely to occur at hospitals with Magnet designation (odds ratio [OR] = 3.45, P = .006), lower medical complexity (OR = 2.57, P = .04), and higher patient-to-trainee ratios (OR = 2.48, P = .05). On average, trainees spend <8 minutes per hour in DPCT. Variation exists in DPCT between hospitals. A less complex case mix, increased patient volume, and Magnet designation were independently associated with increased DPCT.

A Discounted Fuzzy Relational Clustering of Web Users’ Using Intuitive Augmented Sessions Dissimilarity Metric

Relational fuzzy clustering (RFC) algorithms prove very useful in Web user session clustering because Web user sessions may contain fuzzy, conflicting and imprecise information. Though RFC algorithms are very sensitive to cluster initialization and works only if the numbers of clusters are specified in advance. However, at all times, the prior initialization of a number of clusters is not feasible due to the dynamically evolving nature of user sessions. Therefore, estimating the number of clusters and initializing suitable cluster prototype are a significant performance bottleneck in this method. In this paper, the discounted fuzzy relational clustering (DFRC) algorithm is proposed to address the major constraint of RFC. The DFRC algorithm identifies Web user session clusters from Web server access logs, without initializing the number of clusters and prototypes of initial clusters. The DFRC algorithm works in two stages. In the first stage, DFRC automatically identifies the number of potential clusters based on the successively discounted potential density function value of each relational data and their respective centres. In the second stage, DFRC assigns fuzzy membership values to each data point and forms fuzzy clusters from the relational matrix. The DFRC algorithm is applied on an augmented session dissimilarity matrix obtained from a publicly accessed NASA Web server log data. The experimental results are evaluated using different fuzzy validity measures. The extensive experiments are performed to test the effect of various parameters, including accept/reject ratio and neighbourhood radius on the performance of DFRC algorithm. The results were also compared with fuzzy relational clustering algorithm using cluster quality measures. It is observed that the quality of generated clusters using DFRC is superior as compared with that of RFC.

Discovery of web usage patterns using fuzzy mountain clustering

Analysis of web server logs of e-business organisations is critical to provide insight into users' web usage behaviour which can assist in designing most attractive websites. In this article, a mountain density function MDF-based fuzzy clustering framework to discover user session clusters from web logs is proposed. Major steps in this framework include web log preprocessing, MDF-based discovery of user session clusters and their validation. To deal with high dimensionality of user sessions, a fuzzy approach for assigning weights to user sessions has been proposed. For the discovery of user session clusters, fuzzy c-means FCM and fuzzy c-medoids FCMed algorithms are explored. Since the selection of suitable initial cluster centres is a big challenge, MDF-based fuzzy c-means MDFCM and fuzzy c-medoids MDFCMed algorithms are proposed to overcome this problem. Our results show that quality of clusters formed using MDFCM/MDFCMed is much better than FCM and FCMed.

An Efficient Agglomerative Clustering Algorithm for Web Navigation Pattern Identification

Web log mining is analysis of web log files with web page sequences. Discovering user access patterns from web access are necessary for building adaptive web servers, to improve e-commerce, to carry out cross-marketing, for web personalization, to predict web access sequence etc. In this paper, a new agglomerative clustering technique is proposed to identify users with similar interest, and to determine the motivation for visiting a website. Using this approach, web usage mining is done through different stages namely data cleaning, preprocessing, pattern discovery and pattern analysis. Results are given to explain how this approach produces tight usage clusters than the existing web usage mining techniques. Rather than traditional distance based clustering, the similarity measure is considered during clustering process in order to reduce computational complexity. This paper also deals with the problem of assessing the quality of user session clusters and cluster validity is measured by using statistical test, which measures the distances of clusters distributions to infer their dissimilarity and distinguish level. Using such statistical measures, it is proved that cluster accuracy is improved to the extent of 0.83, over existing k-means clustering with validity measure 0.26, FCM (Fuzzy C Means) clustering with validity measure 0.56. Rough set based clustering with validity measure 0.54 Generation of dense clusters is essential for finding interesting patterns needed for further mining and analysis.

A fuzzy neural network based framework to discover user access patterns from web log data

Clustering data from web user sessions is extensively applied to extract customer usage behavior to serve customized content to individual users. Due to the human involvement, web usage data usually contain noisy, incomplete and vague information. Neural networks have the capability to extract embedded knowledge in the form of user session clusters from the huge web usage data. Moreover, they provide tolerance against imperfect and noisy data. Fuzzy sets are another popular tool utilized for handling uncertainty and vagueness hidden in the data. In this paper a fuzzy neural clustering network (FNCN) based framework is proposed that makes use of the fuzzy membership concept of fuzzy c-means (FCM) clustering and the learning rate of a modified self-organizing map (MSOM) neural network model and tries to minimize the weighted sum of the squared error. FNCN is applied to cluster the users’ web access data extracted from the web logs of an educational institution’s proxy web server. The performance of FNCN is compared with FCM and MSOM based clustering methods using various validity indexes. Our results show that FNCN produces better quality of clusters than FCM and MSOM.

Optimization of Clusters of Web Query Sessions using Genetic Algorithm for Effective Personalized Web Search

Personalization of web search is used for effective Information Retrieval in order to better satisfy the information need of the user on the web. The web usage mining has been used widely in Personalization of Web Search(PWS). The effectiveness of the Personalization of Web Search based on clustered web usage data depends on the quality of clusters. It is found in research that there exist no clustering algorithms that produce clusters of 100% quality. In this paper the Genetic Algorithm(GA) is used for clusters optimization in order to improve the quality of clusters for effective Personalized web search. Experiment was conducted on the data set of query sessions captured on the web in Academics, Entertainment and Sports Domain. The search results confirm the improvement in the average precision of the PWS(with cluster optimization) in comparison to PWS( without cluster optimization).