Academic Dataset Research Articles

MCAP: Low-Pass GNNs with Matrix Completion for Academic Recommendations

Graph Neural Networks (GNNs) are commonly used and have shown promising performance in recommendation systems. A major branch, Heterogeneous GNNs, models heterogeneous information by leveraging side information for academic paper recommendations. These networks use message passing and high-order propagation to learn representations for users and items. However, existing recommendation methods perform high-order propagation, leading to suboptimal representation learning. To address this issue, this paper proposes a framework called MCAP, which uses relation-aware GNNs and executes low-pass propagation with matrix completion to enhance academic paper recommendations. The framework uses an attention mechanism to learn top- U relationships by constructing a user-user relation graph based on common authors and venues from interacted items. To efficiently and effectively capture semantic-aware similar items, MCAP builds an item-item relation graph by fusing side information of papers using text embedding models (e.g., Mistral) and large language models (e.g., GPT-3.5-Turbo, GLM-4). Finally, the relation-aware user-user and item-item graphs are incorporated into existing GNN-based models to generate representations of users and papers to enhance academic paper recommendations. The effectiveness of the MCAP is validated using four academic datasets, AMiner-PC, AMiner-WeChat, CiteULike, and DBLP, with user-item interactions and side information of papers. Comprehensive experiments show that the MCAP outperforms state-of-the-art models in terms of Recall@5, NDCG@5, and HR@5 with 69.2%, 70.5%, and 77.6% on the AMiner-WeChat dataset. The code for MCAP is available at https://github.com/THUDM/MCAP .

ICA-CRMAS: Intelligent Context-Awareness Approach for Citation Recommendation based on Multi-Agent System

Navigating the ever-expanding sea of scientific literature presents a daunting challenge for researchers seeking relevant and up-to-date information. Traditional citation recommendation systems, while well-intentioned, often fall short due to their limited focus on text-based features and lack of contextual awareness. In this paper we introduce the ICA-CRMAS (Intelligent Context-Aware Approach for Citation Recommendation based on Multi-Agent System), an intelligent system that leverages the power of deep learning, semantic analysis, and multimodal learning to overcome these limitations. ICA-CRMAS goes beyond the surface, delving into the rich tapestry of information within academic papers, including figures, which often hold vital contextual clues. By weaving this contextual data directly into its recommendation models, ICA-CRMAS generates highly personalized and relevant suggestions. This comprehensive approach unlocks enhanced accuracy, diversity, and serendipity, enabling researchers to effectively discover papers aligning with their interests and research objectives. ICA-CRMAS illuminates its reasoning. Instead of opaque suggestions, the system provides clear explanations that justify and illustrate recommended citations. This transparency builds user confidence, allowing researchers to critically engage with and trust the system’s recommendations. Evaluation experiments conducted on real-world academic datasets demonstrate that ICA-CRMAS outperforms existing approaches across various metrics. it surpassing its closest competitor by a margin of 7.53 on accuracy, 6.07% on MRR and by 5.87 on Recall. User feedback further reinforces its effectiveness, with an Overall System Usability Scale (SUS) score of 76.73, exceeding benchmark scores for comparable systems.

A Military Audio Dataset for Situational Awareness and Surveillance

Audio classification related to military activities is a challenging task due to the high levels of background noise and the lack of suitable and publicly available datasets. To bridge this gap, this paper constructs and introduces a new military audio dataset, named MAD, which is suitable for training and evaluating audio classification systems. The proposed MAD dataset is extracted from various military videos and contains 8,075 sound samples from 7 classes corresponding to approximately 12 hours, exhibiting distinctive characteristics not presented in academic datasets typically used for machine learning research. We present a comprehensive description of the dataset, including its acoustic statistics and examples. We further conduct a comprehensive sound classification study of various deep learning algorithms on the MAD dataset. We are also releasing the source code to make it easy to build these systems. The presented dataset will be a valuable resource for evaluating the performance of existing algorithms and for advancing research in the field of acoustic-based hazardous situation surveillance systems.

Enhancing Academic Integrity: Plagiarism Detection with Machine Learning

This project presents an advanced machine learning-based system designed to empower students to uphold academic integrity and foster originality in their research endeavours. Seamlessly integrated into the research process, the system enables students to input project information, utilizing a sophisticated algorithm and vast academic dataset to compute a matching percentage, facilitating the assessment of work uniqueness. Beyond mere plagiarism detection, the tool serves as an educational resource, promoting self-awareness and improvement among students. It encourages responsible research practices, offers comprehensive reports, and personalized recommendations, and integrates with existing learning management systems for accessibility. Addressing evolving research methodologies, promoting interdisciplinary exploration, and maintaining adaptability to shifting trends, the system significantly contributes to academic integrity preservation and originality cultivation. Key Words: academic integrity, originality, machine learning, plagiarism detection, educational resources, research methodologies.

An academic recommender system on large citation data based on clustering, graph modeling and deep learning

Recommendation (recommender) systems (RS) have played a significant role in both research and industry in recent years. In the area of academia, there is a need to help researchers discover the most appropriate and relevant scientific information through recommendations. Nevertheless, we argue that there is a major gap between academic state-of-the-art RS and real-world problems. In this paper, we present a novel multi-staged RS based on clustering, graph modeling and deep learning that manages to run on a full dataset (scientific digital library) in the magnitude of millions users and items (papers). We run several tests (experiments/evaluation) as a means to find the best approach regarding the tuning of our system; so, we present and compare three versions of our RS regarding recall and NDCG metrics. The results show that a multi-staged RS that utilizes a variety of techniques and algorithms is able to face real-world problems and large academic datasets. In this way, we suggest a way to close or minimize the gap between research and industry value RS.

Discover Frequent Patterns from Academic Data Of Student Information System

Numerous researchers have explored the realm of data mining in education. The primary goal is knowledge discovery, aiming to support staff in efficiently managing educational units, refining student activities, and ultimately elevating learning outcomes. In this study, we utilize association rules mining, implementing the Apriori algorithm to extract insights from academic datasets sourced from the student information system of Sebha University, Libya. Genuine data is sourced from the cloud server. The algorithm is then applied to unveil relationships among 11 attributes within students' academic records spanning four years. The resulting patterns undergo experimental evaluation, considering support and confidence values. These specific rules are subsequently categorised into four classes and scrutinised for further validation. The proposed method yields valuable patterns pertaining to students' academic progress and retains crucial insights for predicting decisions regarding course additions and drops.

Utilizing J48 Algorithm in Predicting Students Dropout in Higher Education Institution

Dropout refers to students who voluntarily withdraw from a course or program prior to completion. University dropouts continue to be a major concern for educators and represent a substantial loss of human resources for society. At Cebu Technological University, it is always a challenge of the Department Chairperson the declining student population, which resulted in the reduction of the number of sections per year level and under loading of faculty.
 This study centers on the creation of a student-dropout model that predicts a student's behavior toward his studies. This model utilized the J48 decision tree algorithm, which extract data from the Student Information System (SIS) portal of the existing institution. Nine hundred sixty-one (961) demographic and academic datasets from students enrolled in the two programs under the College of Computer, Information, and Communications Technology (CCICT) of Cebu Technological University (CTU) with nineteen (19) attributes. During the testing procedure, 10-fold cross-validation was utilized. The J48 pruned tree utilized an average of 3 foliage with 4 as the measure of the tree. The Kappa statistic yields a value of 0.8617 and its Correctly Classified Instances rate of 93.7%. This algorithm helps a lot to the institution in reducing the escalation of the attrition rate and providing proactive measures to address the issue.

TCRec: A novel paper recommendation method based on ternary coauthor interaction

With the explosion in the number of papers being generated, it has become a common practice to recommend papers related to research interests of authors. Existing methods mainly focus on using binary author–paper interactions to mine the research interests of authors. However, sparse binary interactions make it extremely difficult to distinguish the authors’ research interests. This results in an inability to find papers that authors need. Generally, the research interests of coauthors are highly correlated and this is useful in broadening the distinctions between authors’ research interests. Therefore, we propose a new ternary coauthor recommendation (TCRec) method for paper recommendation. Specifically, we construct a coauthor ternary to describe the coauthorship between authors to explore potential ternary interactions. First, an attention-based bidirectional long short-term memory (Bi-LSTM) is built to learn title and abstract information to represent the research area of a paper and accordingly initialize the research interests for each author. Then, an academic heterogeneous graph with a dual attention mechanism is designed to aggregate the semantic information of neighboring nodes and metapaths. This mechanism can be employed to aggregate the relationships between authors and papers in the academic network to describe the research interests for authors. Finally, after constructing the ternary, a multilayer perceptron (MLP) is applied to mine potential associations between the ternary to enrich author–paper interactions for paper recommendation. Extensive experimental results from two real academic datasets not only show the superior performance of our model over state-of-the-art approaches but also demonstrate its potential value for paper recommendation.

Opt-RNN-DBFSVM: Optimal recurrent neural network density based fuzzy support vector machine

Two major problems are encountered when using fuzzy SVM: (a) the number of local minima increases exponentially with the number of samples and (b) the quantity of required computer storage, required for a regular quadratic programming solver, increases by an exponential magnitude as the problem size expands. The Kernel-Adatron family of algorithms gaining attention lately which has allowed to handle very large classification and regression problems. However, these methods treat different types of samples (Noise, border, and core) with the same manner, which causes searches in unpromising areas and increases the number of iterations. In this work, we introduce a hybrid method to overcome these shortcoming, namely Optimal Recurrent Neural Network Density Based fuzzy Support Vector Machine (Opt-RNN-DBFSVM). This method consists of four steps: (a) characterization of different samples, (b) elimination of samples with a low probability of being a support vector, (c) construction of an appropriate recurrent neural network based on an original energy function, and (d) solution of the system of differential equations, managing the dynamics of the RNN, using the Euler–Cauchy method involving an optimal time step. Thanks to its recurrent architecture, the RNN remembers the regions explored during the search process. We demonstrated that RNN-FSVM converges to feasible support vectors and Opt-RNN-DBFSVM has a very low time complexity compared to RNN-FSVM with constant time step, and KAs-FSVM. Several experiments were performed on academic data sets. We used several classification performance measures to compare Opt-RNN-DBFSVM to different classification methods and the results obtained show the good performance of the proposed method.

A Regression Model and a Combination of Academic and Non-Academic Features to Predict Student Academic Performance

Predicting academic performance provides an effective way for students and faculties to monitor their academic progress. The identification of the most significant features was a key outcome of this research, and the college/university databases from online learning platforms are the main academic data sets used to ascertain performance. However, previous research emphasized the addition of other significant features in the prediction of academic performance. Universities’ organizational features include non-academic essential data used in determining student success, but no research has utilized this data to predict student academic performance. Generally, to evaluate binary classification, the relationship between the predicted classifications and the true classifications is analyzed, this approach can lead to the loss of important information from the data. Furthermore, to avoid such loss, this research proposes a regression model by comparing six regression algorithms, and combining academic and non-academic features for prediction student academic performance. Among the various models examined, the gradient-boosted trees regression model demonstrated the lowest error rate. The proposed features have been observed to have a significant impact on academic performance, making them suitable for use in predictions.

Face Anti-spoofing and Liveliness detection using Mobile net and Haar cascade Algorithm

A common biometric approach is another way of face recognition techniques. The method is convenient, easy access to the user and direct comparison with other methods of face recognition due to its rapid development in recent years. Resisting spoofing attacks made on face recognition systems requires Face Anti-Spoofing Systems (FAS) techniques. FAS based on deep learning perform exceptionally well and dominates this area with the emergence of large-scale academic datasets in recent years. The data requirements for training effective anti-spoofing models in this field, however, are large, and there is no way to perform live spoofing. Our paper proposes a combined method of face liveliness detection using Haar-Cascade algorithms and mobile-net classifiers. As a contribution to stimulating future research, we present an overview of technique like deep learning-based FASs. It covers numerous novel and insightful Anti-spoofing approach structured using the modules,1) Eye opening action evaluates with the blinking eye systems and 2) mobile-net classifier module which makes use of a pre-trained version. We wrap up our study with gradually merged these two modules and added them to a basic facial recognition system. Software Python-based results comparisons between classifiers are used to explain efficiency of the suggested approach.

Characterizing the Impact of Data-Damaged Models on Generalization Strength in Intrusion Detection

Generalization is a longstanding assumption in articles concerning network intrusion detection through machine learning. Novel techniques are frequently proposed and validated based on the improvement they attain when classifying one or more of the existing datasets. The necessary follow-up question of whether this increased performance in classification is meaningful outside of the dataset(s) is almost never investigated. This lacuna is in part due to the sparse dataset landscape in network intrusion detection and the complexity of creating new data. The introduction of two recent datasets, namely CIC-IDS2017 and CSE-CIC-IDS2018, opened up the possibility of testing generalization capability within similar academic datasets. This work investigates how well models from different algorithmic families, pretrained on CICIDS2017, are able to classify the samples in CSE-CIC-IDS2018 without retraining. Earlier work has shown how robust these models are to data reduction when classifying state-of-the-art datasets. This work experimentally demonstrates that the implicit assumption that strong generalized performance naturally follows from strong performance on a specific dataset is largely erroneous. The supervised machine learning algorithms suffered flat losses in classification performance ranging from 0 to 50% (depending on the attack class under test). For non-network-centric attack classes, this performance regression is most pronounced, but even the less affected models that classify the network-centric attack classes still show defects. Current implementations of intrusion detection systems (IDSs) with supervised machine learning (ML) as a core building block are thus very likely flawed if they have been validated on the academic datasets, without the consideration for their general performance on other academic or real-world datasets.

MONITORING AND REMOVAL OF FAKE PRODUCT REVIEW USING MACHINE LEARNING (ML)

The employment of Natural Language Processing (NLP) as a means for identifying and eliminating counterfeit products from a dataset is a significant endeavor. The current project employs a Machine Learning (ML) Algorithm to discern fraudulent customer reviews within a given dataset. The algorithm leverages predictive modeling techniques to estimate the authenticity of such reviews, and subsequently assess their validity. The prominence of product reviews as a critical determinant of consumer purchasing behavior in the realm of electronic commerce has led to a concerning escalation in the frequency of counterfeit reviews on websites and applications. It is imperative for the prominent Ecommerce firms to confront the issue of counterfeit product reviews, prior to engaging in the procurement of commodities from established entities. The utilization of this particular approach facilitates the rectification process concerning counterfeit product evaluations, thereby eliminating the prevalence of spammers. This proactive measure works to ensure that there is no potential compromise of trust among users of Ecommerce platforms. Through implementation of this initiative, the enterprise's administration can identify spurious evaluations and subsequently implement appropriate measures directed towards their resolution. The present model has been constructed employing the Naïve Bayes Algorithmic approach. By implementing the algorithm, it is possible to discern spam reviews from those that are legitimate, on websites or applications. To tally the individuals who engage in fraudulent online practices, a dataset is necessary. In our study, the "Amazon Academic Dataset" serves as the fitting reference point to prepare the model and can potentially increase the precision and adaptability of the results.

ProbSAP: A comprehensive and high-performance system for student academic performance prediction

The student academic performance prediction is becoming an indispensable service in the computer supported intelligent education system. But conventional machine learning-based methods can only exploit the sparse discriminative features of student behaviors in imbalanced academic datasets to predict student academic performance (SAP). Furthermore, there is a lack of imbalanced data processing mechanisms that can efficiently capture student characteristics and achievement. Therefore, we propose a comprehensive and high-performance prediction framework to probe SAP characteristics (ProbSAP) on massive educational data, which can resolve imbalanced data issue and improve academic prediction performance for making course final mark prediction. It consists of three main components: collaborative data processing module for enhancing the data quality, scalable metadata clustering module for alleviating the imbalance of academic features, and XGBoost-enhanced SAP prediction module for academic performance forecasting. The collaborative data processing module integrates multi-dimensional academic data, which sustains a good supply for clustering and modeling in the ProbSAP framework. The comparative evaluation results demonstrate that ProbSAP delivers superior accuracy and efficiency improvement for the course final mark prediction of college students over other state-of-the-art methods such as CNN, SVR, RFR, XGBoost, Catboost-SHAP, and AS-SAN. On average, ProbSAP reduces the mean absolute error (MAE) by 84.76%, 72.11%, and 66.49% compared with XGBoost, Catboost-SHAP, and AS-SAN, respectively. It also leads to a better out-sample fit that minimizes prediction errors between 1% and 9% with over 98% of actual samples.

The Usage of Electronic Academic Database Resources Among Lecturers and Postgraduate Students in Western Delta University, Oghara, Delta State, Nigeria

This study investigated the usage of electronic academic databases resources by lecturers and postgraduate students in Western Delta University, Oghara, Delta State, Nigeria. Four objectives with conforming research questions guided the study. The study adopted a descriptive research design. The population of the study consisted of 134 lecturers and postgraduate students in WDU. The instrument for data collection was a structured questionnaire with the title “Usage of Electronic Academic Database Resources Questionnaire (UEADRQ)”. Data was analyzed using frequency counts and percentages. The findings showed that there was a high level of awareness of the electronic academic database by the lecturers and postgraduate students. JSTOR, Elsevier, DOAJ, ProQuest, Science direct, and LexisNexis where been put to maximum usage. Meanwhile, NUC Virtual library, HINARI, Research4life, AGORA and EBSCO Host were used least frequently. The purposes of using the electronic academic datasets were for research, writing of seminar/conference paper, assignments/seminar presentations and getting up-to-date information with subject areas. Lack of search skill, slow internet bandwidth and incessant power outrage were some of the problems preventing against the effective usage of electronic academic database resources in WDU. Based on the findings of this study, it was concluded that electronic academic databases are vital information resources that need to be readily available in academic libraries for use by lecturers and students in the pursuit of their academic advancement. This study recommended that orientation, workshop and seminar should be organized by librarians and university authorities on how to acquire the needed search skills in order to be able to make adequate use of these electronic academic database resources in order to advance academic excellence.

Building Predictive Model by Using Data Mining and Feature Selection Techniques on Academic Dataset

Building Predictive Model by Using Data Mining and Feature Selection Techniques on Academic Dataset

Cytogenetic and developmental toxicity of bisphenol A and bisphenol S in Arbacia lixula sea urchin embryos

Bisphenol S (BP-S) is one of the most important substitutes of bisphenol A (BP-A), and its environmental occurrence is predicted to intensify in the future. Both BP-A and BP-S were tested for adverse effects on early life stages of Arbacia lixula sea urchins at 0.1 up to 100 µM test concentrations, by evaluating cytogenetic and developmental toxicity endpoints. Embryonic malformations and/or mortality were scored to determine embryotoxicity (72 h post-fertilization). It has been reported in academic dataset that bisphenols concentration reached μg/L in aquatic environment of heavily polluted areas. We have chosen concentrations ranging from 0.1–100 μM in order to highlight, in particular, BP-S effects. Attention should be paid to this range of concentrations in the context of the evaluation of the toxicity and the ecological risk of BP-S as emerging pollutant. Cytogenetic toxicity was measured, using mitotic activity and chromosome aberrations score in embryos (6 h post-fertilization). Both BP-A and BP-S exposures induced embryotoxic effects from 2.5 to 100 µM test concentrations as compared to controls. Malformed embryo percentages following BP-A exposure were significantly higher than in BP-S-exposed embryos from 0.25 to 100 µM (with a ~5-fold difference). BP-A, not BP-S exhibited cytogenetic toxicity at 25 and 100 µM. Our results indicate an embryotoxic potential of bisphenols during critical periods of development with a potent rank order to BP-A vs. BP-S. Thus, we show that BP-A alternative induce similar toxic effects to BP-A with lower severity.

Dataset Search and Augmentation

Data has become an indispensable part of our life. However, current mainstream commercial search engines do not support specialized functions for dataset search. A dataset usually consists of both metadata and data content. Existing information retrieval models designed for Web search cannot efficiently extract semantic information inside structured datasets, even when they contain textual content. Developing new algorithms for next-generation search engines to efficiently find datasets can benefit data practitioners in their data discovery experience. In this dissertation, we consider how to effectively perform dataset search and augmentation. We start by providing an end-to-end description of a dataset search engine following the lifecycle of datasets. Our review includes web dataset acquisition techniques, dataset profiling and augmentation methods, and dataset search tasks and corresponding methods. In order to extract datasets from research articles, we present an information extraction framework to determine triples of interest which can be used for academic dataset search. We propose a feature-based method to augment tabular datasets with additional schema labels to help users and systems to better understand the datasets. We develop three methods for tabular dataset search: the first utilizes generated schema labels to enhance the search results; the second adopts pretrained language models to learn matching features; the third models the complex relations in the datasets as one or more graphs and uses graph neural networks to learn representations of queries and tables. To support dataset search in which a query is also a dataset, we propose universal dataset encoders which regard a dataset as a point set so that the encoded dataset representations can be used to search for similar datasets. Extensive experiments across multiple tasks demonstrate the superiority of our proposed methods over the state of the art. Awarded by: Lehigh University, Bethlehem, USA on 10 May 2022. Supervised by: Brian D. Davison. Available at: https://github.com/Zhiyu-Chen/Dissertation/blob/main/Dissertation_Dataset_Search.pdf.

Mortality, disease progression, and disease burden of acute kidney injury in alcohol use disorder subpopulation

BackgroundThe aim of the study was to quantify the relationship between acute kidney injury (AKI) and alcohol use disorder (AUD). MethodsWe used a large academic medical center and the MIMIC-III databases to quantify AKI disease and mortality burden as well as AKI disease progression in the AUD and non-AUD subpopulations. We used the MIMIC-III dataset to compare two different methods of encoding AKI: ICD-9 codes, and the Kidney Disease: Improving Global Outcomes scheme (KDIGO) definition. In addition to the AUD subpopulation, we also present analyses for the hepatorenal syndrome (HRS) and alcohol-related cirrhosis subpopulations identified via ICD-9/ICD-10 coding. ResultsIn both the ICD-9 and KDIGO encodings of AKI, the AUD subpopulation had a higher incidence of AKI (ICD-9: 43.3% vs. 37.92% AKI in the non-AUD subpopulations; KDIGO: 48.65% vs. 40.53%) in the MIMIC-III dataset. In the academic dataset, the AUD subpopulation also had a higher incidence of AKI than the non-AUD subpopulation (ICD-9/ICD-10: 12.76% vs. 10.71%). The mortality rate of the subpopulation with both AKI and AUD, HRS, or alcohol-related cirrhosis was consistently higher than that of the subpopulation with only AKI in both datasets, including after adjusting for disease severity using two methods of severity estimation in the MIMIC-III dataset. Disease progression rates were similar for AUD and non-AUD subpopulations. ConclusionsOur work shows that the AUD patient subpopulation had a higher number of AKI patients than the non-AUD subpopulation, and that patients with both AKI and AUD, HRS, or alcohol-related cirrhosis had higher rates of mortality than the non-AUD subpopulation with AKI.

Deep Learning for Face Anti-Spoofing: A Survey.

Face anti-spoofing (FAS) has lately attracted increasing attention due to its vital role in securing face recognition systems from presentation attacks (PAs). As more and more realistic PAs with novel types spring up, early-stage FAS methods based on handcrafted features become unreliable due to their limited representation capacity. With the emergence of large-scale academic datasets in the recent decade, deep learning based FAS achieves remarkable performance and dominates this area. However, existing reviews in this field mainly focus on the handcrafted features, which are outdated and uninspiring for the progress of FAS community. In this paper, to stimulate future research, we present the first comprehensive review of recent advances in deep learning based FAS. It covers several novel and insightful components: 1) besides supervision with binary label (e.g., '0' for bonafide versus '1' for PAs), we also investigate recent methods with pixel-wise supervision (e.g., pseudo depth map); 2) in addition to traditional intra-dataset evaluation, we collect and analyze the latest methods specially designed for domain generalization and open-set FAS; and 3) besides commercial RGB camera, we summarize the deep learning applications under multi-modal (e.g., depth and infrared) or specialized (e.g., light field and flash) sensors. We conclude this survey by emphasizing current open issues and highlighting potential prospects.