On Developing Generic Models for Predicting Student Outcomes in Educational Data Mining

Poor academic performance of students is a concern in the educational sector, especially if it leads to students being unable to meet minimum course requirements. However, with timely prediction of students’ performance, educators can detect at-risk students, thereby enabling early interventions for supporting these students in overcoming their learning difficulties. However, the majority of studies have taken the approach of developing individual models that target a single course while developing prediction models. These models are tailored to specific attributes of each course amongst a very diverse set of possibilities. While this approach can yield accurate models in some instances, this strategy is associated with limitations. In many cases, overfitting can take place when course data is small or when new courses are devised. Additionally, maintaining a large suite of models per course is a significant overhead. This issue can be tackled by developing a generic and course-agnostic predictive model that captures more abstract patterns and is able to operate across all courses, irrespective of their differences. This study demonstrates how a generic predictive model can be developed that identifies at-risk students across a wide variety of courses. Experiments were conducted using a range of algorithms, with the generic model producing an effective accuracy. The findings showed that the CatBoost algorithm performed the best on our dataset across the F-measure, ROC (receiver operating characteristic) curve and AUC scores; therefore, it is an excellent candidate algorithm for providing solutions on this domain given its capabilities to seamlessly handle categorical and missing data, which is frequently a feature in educational datasets.

Metapragmatics and genre: Connecting the strands

Genre analysis involves at least a foray into the social/contextual dimension framing genre-exemplars. One way to explore this dimension is drawing on the concept of metapragmatics, which is primarily associated with (American) linguistic anthropology. However, with a few exceptions, genre studies have not consistently operationalized metapragmatics, either theoretically or practically. The purpose of this article is, therefore, to explore one possible angle of such operationalization by means of studying discourse fragments reflecting on fragments of (these very or other) discourses (so-called metapragmatic discourses) vis--vis any generic properties of the reflected discourse. Specifically, we analyzed comments sections for a number of YouTube videos exemplifying several lifestyle genres. The results indicate that generic references can range from simply using a generic label to refer to the discourse in question (as a token of a certain type/genre) to actually discussing the generic characteristics of the genre it instantiates, as well as projecting certain (generic) metapragmatic stances. Another observation is that different wordings used by the discourse community to refer to generic models can be, as it were, proper generic labels, but they can also be words and phrases that would hardly qualify as proper names of genres from an analysts point of view. Both these proper and other - genre-like - labels are also often used in conjunction with or are replaced by other ways of metapragmatically referring to what the speaker does or even what they are in/by dint of using the discourse in question. This suggests that any generic labels or cues are just part of a large pool of other possible metapragmatic meanings, knowledge, and ideologies circulating in discourse communities. More broadly, the results may indicate that genre studies should see genre as an even less stabilized entity because what a genre is depends on what people who actually use it make of it, as well as augment their standard toolkits with methods aimed at exploring metapragmatic discourse.

Generalisation Gap of Keyword Spotters in a Cross-Speaker Low-Resource Scenario

Models for keyword spotting in continuous recordings can significantly improve the experience of navigating vast libraries of audio recordings. In this paper, we describe the development of such a keyword spotting system detecting regions of interest in Polish call centre conversations. Unfortunately, in spite of recent advancements in automatic speech recognition systems, human-level transcription accuracy reported on English benchmarks does not reflect the performance achievable in low-resource languages, such as Polish. Therefore, in this work, we shift our focus from complete speech-to-text conversion to acoustic similarity matching in the hope of reducing the demand for data annotation. As our primary approach, we evaluate Siamese and prototypical neural networks trained on several datasets of English and Polish recordings. While we obtain usable results in English, our models’ performance remains unsatisfactory when applied to Polish speech, both after mono- and cross-lingual training. This performance gap shows that generalisation with limited training resources is a significant obstacle for actual deployments in low-resource languages. As a potential countermeasure, we implement a detector using audio embeddings generated with a generic pre-trained model provided by Google. It has a much more favourable profile when applied in a cross-lingual setup to detect Polish audio patterns. Nevertheless, despite these promising results, its performance on out-of-distribution data are still far from stellar. It would indicate that, in spite of the richness of internal representations created by more generic models, such speech embeddings are not entirely malleable to cross-language transfer.

Best Practice Methodology to Enhance Modeling Inflow Control Technologies in Dynamic Reservoir Simulators

The methodology presented here will expand on current modeling of Autonomous Inflow Control Devices (AICD) to generalize for a wider range of fluid flow rates and phases. It will also address the challenges of modeling multiphase behavior of the reservoir fluid flow. This paper presents proposed methods for selected devices, and device models supported by simulations. The proposed methods show the potential for qualified benchmarking of Inflow Control Technology (ICT) completed wells in dynamic reservoir simulations compared to the generic models currently in use. New single-phase models for segregated and sequential flow are presented, and these have a potential for greatly simplifying mass flowrate predictions for multi-phase flow leading to more accurate analysis within dynamic reservoir simulators.

The comparison and analysis of Type 3 wind turbine models used for researching the stability of electric power systems

The dominant trend of the modern energy is the use of generating plants based on renewable energy sources, among which the most common is a wind power plant based on doubly fed induction generator (Type 3 WT). The large-scale introduction of Type 3 WT into the modern power systems significantly changes their dynamic properties. There are problems with ensuring the basic condition of the reliability and the survivability of power systems – the stability. The study and solution of the indicated problems is possible only with the help of the mathematical modeling of a large-scale power systems which is currently being carried out with the help of widespread purely numerical software tools of calculations of modes and processes, which are characterized by various simplifications and limitations. For the properties and capabilities of software tools for studying stability issues, mathematical models of Type 3 WT, the so-called generic models, which also have simplifications and limitations, are adapted. In this article, the reliability of stability calculations of a real power system with Type 3 WT using software tools was evaluated, which allows to identify the influence of the applied simplifications and restrictions with a purely numerical approach on the quality of solving problems of assessing the stability of power systems with Type 3 WT. Also, the studies made it possible to identify the areas of the application of generic models of Type 3 WT as a part of the model of the real dimension power system, at which the greatest and least errors arise, as well as their causes. Such a comprehensive assessment becomes feasible due to the alternative approach proposed in the article, based on the use of a detail benchmark tool model instead of the full-scale data to compare the results of modeling.

Predicting compound-protein interaction using hierarchical graph convolutional networks

MotivationConvolutional neural networks have enabled unprecedented breakthroughs in a variety of computer vision tasks. They have also drawn much attention from other domains, including drug discovery and drug development. In this study, we develop a computational method based on convolutional neural networks to tackle a fundamental question in drug discovery and development, i.e. the prediction of compound-protein interactions based on compound structure and protein sequence. We propose a hierarchical graph convolutional network (HGCN) to encode small molecules. The HGCN aggregates a molecule embedding from substructure embeddings, which are synthesized from atom embeddings. As small molecules usually share substructures, computing a molecule embedding from those common substructures allows us to learn better generic models. We then combined the HGCN with a one-dimensional convolutional network to construct a complete model for predicting compound-protein interactions. Furthermore we apply an explanation technique, Grad-CAM, to visualize the contribution of each amino acid into the prediction.ResultsExperiments using different datasets show the improvement of our model compared to other GCN-based methods and a sequence based method, DeepDTA, in predicting compound-protein interactions. Each prediction made by the model is also explainable and can be used to identify critical residues mediating the interaction.Availability and implementationhttps://github.com/banhdzui/cpi_hgcn.git

Novel generic models for differentiating stem cells reveal oscillatory mechanisms

Understanding cell fate selection remains a central challenge in developmental biology. We present a class of simple yet biologically motivated mathematical models for cell differentiation that generically generate oscillations and hence suggest alternatives to the standard framework based on Waddington’s epigenetic landscape. The models allow us to suggest two generic dynamical scenarios that describe the differentiation process. In the first scenario, gradual variation of a single control parameter is responsible for both entering and exiting the oscillatory regime. In the second scenario, two control parameters vary: one responsible for entering, and the other for exiting the oscillatory regime. We analyse the standard repressilator and four variants of it and show the dynamical behaviours associated with each scenario. We present a thorough analysis of the associated bifurcations and argue that gene regulatory networks with these repressilator-like characteristics are promising candidates to describe cell fate selection through an oscillatory process.

Biomass models for three species with different growth forms and geographic distribution in the Brazilian Atlantic forest

Allometric models embedding independent variables such as diameter at breast height (d) and total height (h) are useful tools to predict the biomass of individual trees. Models for tropical forests are often constructed based on datasets composed of species with different morphological features and architectural models. It is reasonable to expect, however, that species-specific models may reduce uncertainties in biomass predictions, especially for palms, tree ferns, and trees with peculiar morphological features, such as stilt roots and hollow trunks. In this sense, three species with wide geographical distribution in the Brazilian Atlantic Forest were sampled, namely Euterpe edulis Mart., Cyathea delgadii Sternb., and Cecropia glaziovii Snethl., with the aim to (i) quantify their aboveground biomass (AGB), (ii) evaluate the AGB distribution in different plant compartments, (iii) fit species-specific models for predicting AGB at the individual level, and (iv) assess the performance of specific and generic models available in the literature to predict the AGB of individuals of these species. The compartment stem represented, on average, ∼74% of the total AGB of E. edulis individuals; in turn, the caudex compartment of C. delgadii represented, on average, ∼87% of the total AGB, while the trunk compartment of C. glaziovii represented, on average, ∼74%. Among the fitted models, the power model [Formula: see text] showed the best performance for E. edulis and C. delgadii. In turn, the asymptotic logistic model [Formula: see text], where dc is the diameter above the upper stilt root, presented the best performance for C. glaziovii. The variable h appeared as the most important predictor of AGB of E. edulis and C. delgadii; in contrast, the stem and caudex mean basic specific gravities were not suitable predictors. The fitted species-specific models outperformed the specific and generic models selected from the literature. They may, therefore, contribute to the reduction of uncertainties in AGB estimates. In addition, the results support evidence that specific models may be necessary for species with different growth forms and (or) peculiar morphological features, especially those with great abundance and wide geographic distribution.