Generic Predictive Model Research Articles

AlzDiscovery: A computational tool to identify Alzheimer's disease-causing missense mutations using protein structure information.

Alzheimer's disease (AD) is one of the most common forms of dementia and neurodegenerative diseases, characterized by the formation of neuritic plaques and neurofibrillary tangles. Many different proteins participate in this complicated pathogenic mechanism, and missense mutations can alter the folding and functions of these proteins, significantly increasing the risk of AD. However, many methods to identify AD-causing variants did not consider the effect of mutations from the perspective of a protein three-dimensional environment. Here, we present a machine learning-based analysis to classify the AD-causing mutations from their benign counterparts in 21 AD-related proteins leveraging both sequence- and structure-based features. Using computational tools to estimate the effect of mutations on protein stability, we first observed a bias of the pathogenic mutations with significant destabilizing effects on family AD-related proteins. Combining this insight, we built a generic predictive model, and improved the performance by tuning the sample weights in the training process. Our final model achieved the performance on area under the receiver operating characteristic curve up to 0.95 in the blind test and 0.70 in an independent clinical validation, outperforming all the state-of-the-art methods. Feature interpretation indicated that the hydrophobic environment and polar interaction contacts were crucial to the decision on pathogenic phenotypes of missense mutations. Finally, we presented a user-friendly web server, AlzDiscovery, for researchers to browse the predicted phenotypes of all possible missense mutations on these 21 AD-related proteins. Our study will be a valuable resource for AD screening and the development of personalized treatment.

From a local descriptive to a generic predictive model of cereal aphid regulation by predators.

The temporal dynamics of insect populations in agroecosystems are influenced by numerous biotic and abiotic interactions, including trophic interactions in complex food webs. Predicting the regulation of herbivorous insect pests by arthropod predators and parasitoids would allow for rendering crop production less dependent on chemical pesticides. Curtsdotter etal. (2019) developed a food-web model simulating the influences of naturally occurring arthropod predators on aphid population dynamics in cereal crop fields. The use of an allometric hypothesis based on the relative body masses of the prey and various predator guilds reduced the number of estimated parameters to just five, albeit field-specific. Here, we extend this model and test its applicability and predictive capacity. We first parameterized the original model with a dataset with the dynamic arthropod community compositions in 54 fields in six regions in France. We then integrated three additional biological functions to the model: parasitism, aphid carrying capacity and suboptimal high temperatures that reduce aphid growth rates. We developed a multi-field calibration approach to estimate a single set of generic allometric parameters for a given group of fields, which would increase model generality needed for predictions. The original and revised models, when using field-specific parameterization, achieved quantitatively good fits to observed aphid population dynamics for 59% and 53% of the fields, respectively, with pseudo-R2 up to 0.99. But the multi-field calibration showed that increased model generality came at the cost of reduced model reliability (goodness-of-fit). Our study highlights the need to further improve our understanding of how body size and other traits affect trophic interactions in food webs. It also points up the need to acquire high-resolution data to use this type of modelling approach. We propose that a hypothesis-driven strategy of model improvement based on the integration of additional biological functions and additional functional traits beyond body size (e.g., predator space search or prey defences) into the food-web matrix can improve model reliability.

ProkDBP: Toward more precise identification of prokaryotic DNA binding proteins.

Prokaryotic DNA binding proteins (DBPs) play pivotal roles in governing gene regulation, DNA replication, and various cellular functions. Accurate computational models for predicting prokaryotic DBPs hold immense promise in accelerating the discovery of novel proteins, fostering a deeper understanding of prokaryotic biology, and facilitating the development of therapeutics targeting for potential disease interventions. However, existing generic prediction models often exhibit lower accuracy in predicting prokaryotic DBPs. To address this gap, we introduce ProkDBP, a novel machine learning-driven computational model for prediction of prokaryotic DBPs. For prediction, a total of nine shallow learning algorithms and five deep learning models were utilized, with the shallow learning models demonstrating higher performance metrics compared to their deep learning counterparts. The light gradient boosting machine (LGBM), coupled with evolutionarily significant features selected via random forest variable importance measure (RF-VIM) yielded the highest five-fold cross-validation accuracy. The model achieved the highest auROC (0.9534) and auPRC (0.9575) among the 14 machine learning models evaluated. Additionally, ProkDBP demonstrated substantial performance with an independent dataset, exhibiting higher values of auROC (0.9332) and auPRC (0.9371). Notably, when benchmarked against several cutting-edge existing models, ProkDBP showcased superior predictive accuracy. Furthermore, to promote accessibility and usability, ProkDBP (https://iasri-sg.icar.gov.in/prokdbp/) is available as an online prediction tool, enabling free access to interested users. This tool stands as a significant contribution, enhancing the repertoire of resources for accurate and efficient prediction of prokaryotic DBPs.

Systematic development and validation of a predictive model for major postoperative complications in the Peri-operative Quality Improvement Project (PQIP) dataset.

Complications are common following major surgery and are associated with increased use of healthcare resources, disability and mortality. Continued reliance on mortality estimates risks harming patients and health systems, but existing tools for predicting complications are unwieldy and inaccurate. We aimed to systematically construct an accurate pre-operative model for predicting major postoperative complications; compare its performance against existing tools; and identify sources of inaccuracy in predictive models more generally. Complete patient records from the UK Peri-operative Quality Improvement Programme dataset were analysed. Major complications were defined as Clavien-Dindo grade ≥ 2 for novel models. In a 75% train:25% test split cohort, we developed a pipeline of increasingly complex models, prioritising pre-operative predictors using Least Absolute Shrinkage and Selection Operators (LASSO). We defined the best model in the training cohort by the lowest Akaike's information criterion, balancing accuracy and simplicity. Of the 24,983 included cases, 6389 (25.6%) patients developed major complications. Potentially modifiable risk factors (pain, reduced mobility and smoking) were retained. The best-performing model was highly complex, specifying individual hospital complication rates and 11 patient covariates. This novel model showed substantially superior performance over generic and specific prediction models and scores. We have developed a novel complications model with good internal accuracy, re-prioritised predictor variables and identified hospital-level variation as an important, but overlooked, source of inaccuracy in existing tools. The complexity of the best-performing model does, however, highlight the need for a step-change in clinical risk prediction to automate the delivery of informative risk estimates in clinical systems.

Generic prediction model of moisture content for maize kernels by combing spectral and color data through hyperspectral imaging

Moisture content (MC) is an important index to measure the quality of maize kernels. This study aims to construct a generic prediction model and optimize the characteristic variables for efficiently predicting the MC of maize kernels across various varieties. Visible/near-infrared hyperspectral imaging (HSI) within the wavelength range of 374.98 - 1038.79 nm is employed. The 270 samples containing 18 varieties at five periods over two years are collected. Spectral and color feature data based on the H, S, and V color channels of maize are extracted. Partial least squares regression (PLSR) and principal component regression (PCR) are used to establish MC prediction models. Backward interval least squares regression (Bi-PLS), uninformative variables elimination (UVE), and successive projections algorithm (SPA) are jointly (Bi-PLS, UVE, Bi-PLS-SPA, UVE-SPA, and SPA) used to select characteristic wavelengths. The MC prediction models are developed using whole wavelengths, characteristic wavelengths, color features, and fused data. The results suggest that the optimal PLSR model is based on fused data of color features and characteristic wavelengths selected by UVE-SPA from S-G smoothing spectral data (HSV+S-G-UVE-SPA-PLSR). The Rc and Rp are 0.9804 and 0.9835, respectively. The RMCEc and RMCEp are 1.6889% and 1.6523%, respectively. The RPD is 5.22. The optimal prediction model can quickly measure MC for different maize kernels, guiding diverse application scenarios to enhance the quality of maize kernels and processing efficiency.

Prediction model for multiblock data with or without intermediate blocks. Estimation by PLS Path Modeling

An important issue in modern science remains the integration of multiple data sources, considered as datasets divided into blocks of variables measured on the same set of observations. When analyzing these blocks, the path modeling approach aims to take into account a specific pattern of directed relations between them. This structure, usually set up from prior knowledge, leads to a path diagram in which the block’s components are linked to one another. Until path modeling has gained popularity in recent decades, the issue of prediction has been surprisingly neglected in the literature. To fill this gap, we propose a prediction framework dedicated to the path modeling approach. Indeed, prediction for path modeling methods is of paramount interest for: (i) selecting the optimal model with cross-validation (e.g., optimal model specification and dimension), (ii) comparing the model with other relevant ones, (iii) or predicting new observations. Here, we propose an explicit formulation of a prediction model in the context of a path modeling structure. This generic prediction model is estimated using the PLS Path Modeling algorithm, denoted PMM-PLSPM (=Prediction Model for Multiblock data estimated by PLS Path Modeling). Nevertheless, its versatility makes it possible to adapt the approach to any other composite-based methods. Two estimation strategies (PLSPM-like and PLSR-like) are proposed for prediction, both being viewed as the structural model at the variable level. At the same time, the components of multiple blocks – obtained by a relevant deflation strategy – are handled to take advantage of the full multidimensional potential of blocks. The proposed prediction strategies are applied and evaluated on real data sets.

Decoupling Analysis of O2 Adsorption on Metal-N-C Single-Atom Catalysts via Data-Driven Descriptors.

The adsorption energy of adsorbed molecules on single-atom catalysts is a key indicator of the catalytic activity of the catalysts. Developing a generic and interpretable structure-property prediction model from numerous influencing factors is a challenging task. In this work, we constructed a machine learning (ML) model from first-principles calculations of the adsorption energy data of O2 on Ni(II), Co(II), Cu(II), Fe(II), Fe(III), and Mn(II) single-atom catalysts supported on 15 different N-C substrates under various spin states. A mathematic formula is proposed to predict the adsorption energy by a novel data-driven descriptor derived from physically meaningful factors such as geometric distances and atomic charges. This data-driven descriptor is relevant to only the geometrical configuration of the adsorbate, while the parameters in the linear formulas contain only substrate-specific information. This ML model with the ability to decouple variables will greatly advance the understanding of metal-N-C single-atom catalysts and help in the design of new substrates to modulate catalytic activity.

Objective Prediction of Next-Day's Affect Using Multimodal Physiological and Behavioral Data: Algorithm Development and Validation Study.

Affective states are important aspects of healthy functioning; as such, monitoring and understanding affect is necessary for the assessment and treatment of mood-based disorders. Recent advancements in wearable technologies have increased the use of such tools in detecting and accurately estimating mental states (eg, affect, mood, and stress), offering comprehensive and continuous monitoring of individuals over time. Previous attempts to model an individual's mental state relied on subjective measurements or the inclusion of only a few objective monitoring modalities (eg, smartphones). This study aims to investigate the capacity of monitoring affect using fully objective measurements. We conducted a comparatively long-term (12-month) study with a holistic sampling of participants' moods, including 20 affective states. Longitudinal physiological data (eg, sleep and heart rate), as well as daily assessments of affect, were collected using 3 modalities (ie, smartphone, watch, and ring) from 20 college students over a year. We examined the difference between the distributions of data collected from each modality along with the differences between their rates of missingness. Out of the 20 participants, 7 provided us with 200 or more days' worth of data, and we used this for our predictive modeling setup. Distributions of positive affect (PA) and negative affect (NA) among the 7 selected participants were observed. For predictive modeling, we assessed the performance of different machine learning models, including random forests (RFs), support vector machines (SVMs), multilayer perceptron (MLP), and K-nearest neighbor (KNN). We also investigated the capability of each modality in predicting mood and the most important features of PA and NA RF models. RF was the best-performing model in our analysis and performed mood and stress (nervousness) prediction with ~81% and ~72% accuracy, respectively. PA models resulted in better performance compared to NA. The order of the most important modalities in predicting PA and NA was the smart ring, phone, and watch, respectively. SHAP (Shapley Additive Explanations) analysis showed that sleep and activity-related features were the most impactful in predicting PA and NA. Generic machine learning-based affect prediction models, trained with population data, outperform existing methods, which use the individual's historical information. Our findings indicated that our mood prediction method outperformed the existing methods. Additionally, we found that sleep and activity level were the most important features for predicting next-day PA and NA, respectively.

Unified compressive strength model for axially loaded circular composite short columns

Composite columns are widely used in modern structures owing primarily to their excellent structural performance, but they usually have very different ultimate axial compressive strength prediction models, which may lead to confusion for researchers and engineering designers. In addition, given the similarity of their mechanical mechanisms, there may be a generic prediction model for these columns. Hence, the authors aim to provide a simple and effective unified ultimate compression prediction model for axially loaded circular composite short columns. A rich experimental database is firstly established. Then the force analysis of a typical concrete-filled double-skin steel tubes (CFDST) column is illustrated, and it can be easily extended to other types of composite columns. Based on the above force analysis, the stress state of each component of the column can be determined via different sections and loading concepts. Contributions of the inner steel tube to the load-bearing capacity are further clarified, and the passively confined compressive strength of the concrete core is also investigated on the basis of the proposed equivalent confining stress. Subsequently, the calculation flow of the ultimate compression strength is briefly demonstrated. Finally, the ultimate compression prediction model is evaluated by the established test database, and good consistency can be observed for circular composite short columns with a widely varied parameter.

Random forest incorporating ab-initio calculations for corrosion rate prediction with small sample Al alloys data

Corrosion jeopardizes the materials longevity and engineering safety, hence the corrosion rate needs to be forecasted so as to better guide materials selection. Although field exposure experiments are dependable, the prohibitive cost and their time-consuming nature make it difficult to obtain large dataset for machine learning. Here, we propose a strategy Integrating Ab-initio Calculations with Random Forest (IACRF) to optimize the model, thereby estimating the corrosion rate of Al alloys in diverse environments. Based on the thermodynamic assessment of the secondary phases, the ab-initio calculation quantities, especially the work function, significantly improved the prediction accuracy with respect to small-sample Al alloys corrosion dataset. To build a better generic prediction model, the most accessible and effective features are identified to train IACRF. Finally, the independent field exposure experiments in Southeast Asia have proven the generalization ability of IACRF in which the average prediction accuracy is improved up to 91%.

A generic yield prediction model for grapes under agro climatic conditions based on disease management

Climate has a direct influence on crop development and the final yield. In this article, a generic agro climatic yield prediction model for grape is developed and analytically solved. This model is useful for research scholars, faculty members and academicians in the area of mathematical biology. The asymptotic analysis is carried out to obtain the final form of the yield prediction model. In the process of model development, climate, disease and grape yield are considered as dependent parameter. Infection rate, disease incidence, seasonality rate and removal rate of grape yield per harvest time are considered as independent parameters. Further, the model is studied and the parameters estimation from the field level data during the period 2015-2021 from GRS and Theni surrounding villages. The effects of various parameters on concentration curves are discussed. Stability Analysis of this model is also explained. The obtained analytical solution in comparison with the numerical and stability analysis is found to be in satisfactory agreement. In addition, the basic reproduction number for this model is obtained. This model helps to predict the future calculations of infected and recovered yield for grape from the reproduction number R0. The model permits to highlight crucial mechanisms to undergo and evaluate the consequences of different agricultural practices on the quantity and quality of the yield.

BIRAFFE2, a multimodal dataset for emotion-based personalization in rich affective game environments

Generic emotion prediction models based on physiological data developed in the field of affective computing apparently are not robust enough. To improve their effectiveness, one needs to personalize them to specific individuals and incorporate broader contextual information. To address the lack of relevant datasets, we propose the 2nd Study in Bio-Reactions and Faces for Emotion-based Personalization for AI Systems (BIRAFFE2) dataset. In addition to the classical procedure in the stimulus-appraisal paradigm, it also contains data from an affective gaming session in which a range of contextual data was collected from the game environment. This is complemented by accelerometer, ECG and EDA signals, participants’ facial expression data, together with personality and game engagement questionnaires. The dataset was collected on 102 participants. Its potential usefulness is presented by validating the correctness of the contextual data and indicating the relationships between personality and participants’ emotions and between personality and physiological signals.

Generic predictive model calibration for PMSMs with different topologies

The main difficulty in designing predictive controllers for permanent magnet motors with different topologies is the control of the additional current subspaces. The frequency of current components in these additional subspaces (harmonic subspaces and zero-sequence loop/subspaces) is several times the fundamental current frequency. Moreover, the inner voltage source would vary with operation conditions, and there are also different views about the causes of inner voltage sources. This paper analyses the inner voltage sources in these additional current subspaces for permanent magnet motors with different drive topologies using simulation and experiment. Based on the results of the simulations and experiments, we analyze the sources and characteristics of the inner voltage sources and provide a predictive model calibration method for the components in these additional subspaces. Finally, the deadbeat current control scheme based on calibrated look-up table is designed and verified.

Physical Metallurgy Guided Industrial Big Data Analysis System with Data Classification and Property Prediction

Various computational analysis systems based on machine learning (ML) methods have been established for the analysis of steel industrial data. However, limited by the extensibility of one regression strategy, it is difficult to obtain a generic property prediction model for multiple types of steels. To solve this problem, this study proposes a novel industrial big data analysis system that combines ML classification and regression models with key physical metallurgy (PM) variables. First, the database is obtained from an industrial production line and carefully preprocessed. Then, multiple types of steels are categorized into five classes using a K‐nearest neighbor (KNN) algorithm, and suitable ML algorithms are selected for each category to maximize the performance. Considering the role of PM variables in improving the model accuracy, some relevant parameters (the Ac1 temperature, Ac3 temperature, and flow stress) are introduced to guide the further optimization of the ML process. The proposed industrial analysis system has more accurate prediction and higher flexibility than the model that directly uses the original dataset. With a rational combination of different regression strategies, the present results clearly demonstrate that the extensibility of the proposed property prediction model is significantly improved for industrial big data.

Generic Predictive Model of Earthquake-Induced Slope Displacements Derived from Finite-Element Analysis

A generic predictive model of earthquake-induced slope displacement subjected to shallow crustal earthquake events is developed using displacements computed from finite-element (FE) analysis. The maximum displacement on the slope surface at the end of shaking was computed by nonlinear FE simulations for 49 slope models each subjected to 1051 earthquake motions. A unified predictive model of seismic displacement is developed that characterizes the slope in terms of its yield acceleration (ky), the depth of the slip surface relative to the height of the slope (Hratio), and the natural period of the full slope height (Tslope). Across five intensity measures and 10 combinations of intensity measures, peak ground velocity (PGV) is found to be the most efficient parameter for the displacement prediction, leading to significantly smaller aleatory variability. The displacement variability is partitioned into two components: between-slope variability, which represents the variability associated with different slope models, and within-slope variability, which represents the variability due to different input ground motions. The developed generic predictive model can be applied to the probabilistic seismic hazard analysis of slope movements and used for deterministic earthquake scenarios in the design/analysis process.

Kalman filter updating of rutting predictive models in flexible pavements using measured field data

ABSTRACT Predicting pavement rutting is associated with significant uncertainties that often lead to inefficient maintenance planning. The predictive performance of rutting models is exacerbated in local road agencies and developing countries that rely on generic and knowledge-based models which are typically unreliable if used without adaptation, validation, or calibration. This study aims at developing a probabilistic framework that employs Ensemble Kalman Filter (EnKF) techniques to update the parameters associated with generic rutting predictive models while accounting for the prevailing uncertainties. When coupled with a continuous influx of measured data, the EnKF framework sequentially updates the generic models and minimizes prediction errors in real-time. The robustness of the presented scheme is demonstrated through a numerical example, and its sensitivity to the use of different generic curves as starting points is examined. The results indicate that the EnKF framework improves the accuracy of rutting predictions by up to 60% and that accuracy remains within tolerable limits whilst varying the range of the uncertainty in the measurements or the initial states. The paper concludes with a discussion of how practitioners can integrate the outcomes of the presented framework to enact maintenance policies that minimize the financial cost at the project and network levels.

A LUCAS‐based mid‐infrared soil spectral library: Its usefulness for soil survey and precision agriculture#

AbstractBackgroundMid‐infrared spectroscopy (MIRS) is commonly recognized as a rapid and high throughput measurement technology for numerous soil properties, given that appropriate prediction models are calibrated. Soil spectral libraries (SSL) may reduce effort and costs for MIRS practical application.AimsTo calibrate MIR‐SSL‐based prediction models for soil properties and to test their applicability to independent sample sets at regional scale (e.g., for soil survey) and at field scale (i.e., for precision agriculture, PA).MethodsSpectra of 1013 arable topsoil samples of the European Land Use/Land Cover Area Frame Survey 2009 (LUCAS) from Belgium, the Netherlands, Luxembourg, and Germany formed the basis for the MIR‐SSL. Leave‐one‐out cross‐validation (LOOCV) via partial least squares regression served to calibrate (1) generic prediction models including all samples, and (2) stratified models for different parent materials. Test‐set validation (TSV) was conducted on samples from independent campaigns at (1) regional scale with a sample set from Schleswig‐Holstein (Germany; n = 385) and (2) field scale for four individual fields in Germany (n = 513).ResultsGeneric LOOCV models successfully predicted soil organic carbon, total nitrogen, sand, silt, clay, carbonate, and pH. Calibration for available nutrients failed. The TSV was successful for the regional sample set for all variables (2.5 ≤ RPIQ ≤ 5.9), except for carbonate (RPIQ = 0). At field scale, the validation was highly variable for different sites and parameters. Stratified models using soil parent material as auxiliary variable improved only occasionally the applicability at field scale, that is, on single fields and only for clay and carbonate.ConclusionsAlthough the MIR‐SSL in its present state cannot be recommended for nutrient management, it provides valuable support for soil survey and PA.

Population-centric risk prediction modeling for gestational diabetes mellitus: A machine learning approach

AimsThe heterogeneity in Gestational Diabetes Mellitus (GDM) risk factors among different populations impose challenges in developing a generic prediction model. This study evaluates the predictive ability of existing UK NICE guidelines for assessing GDM risk in Singaporean women, and used machine learning to develop a non-invasive predictive model. MethodsData from 909 pregnancies in Singapore’s most deeply phenotyped mother-offspring cohort study, Growing Up in Singapore Towards healthy Outcomes (GUSTO), was used for predictive modeling. We used a CatBoost gradient boosting algorithm, and the Shapley feature attribution framework for model building and interpretation of GDM risk attributes. ResultsUK NICE guidelines showed poor predictability in Singaporean women [AUC:0.60 (95% CI 0.51, 0.70)]. The non-invasive predictive model comprising of 4 non-invasive factors: mean arterial blood pressure in first trimester, age, ethnicity and previous history of GDM, greatly outperformed [AUC:0.82 (95% CI 0.71, 0.93)] the UK NICE guidelines. ConclusionsThe UK NICE guidelines may be insufficient to assess GDM risk in Asian women. Our non-invasive predictive model outperforms the current state-of-the-art machine learning models to predict GDM, is easily accessible and can be an effective approach to minimize the economic burden of universal testing & GDM associated healthcare in Asian populations.

Predicting the objective and priority of issue reports in software repositories

Software repositories such as GitHub host a large number of software entities. Developers collaboratively discuss, implement, use, and share these entities. Proper documentation plays an important role in successful software management and maintenance. Users exploit Issue Tracking Systems, a facility of software repositories, to keep track of issue reports, to manage the workload and processes, and finally, to document the highlight of their team’s effort. An issue report is a rich source of collaboratively-curated software knowledge, and can contain a reported problem, a request for new features, or merely a question about the software product. As the number of these issues increases, it becomes harder to manage them manually. GitHub provides labels for tagging issues, as a means of issue management. However, about half of the issues in GitHub’s top 1000 repositories do not have any labels. In this work, we aim at automating the process of managing issue reports for software teams. We propose a two-stage approach to predict both the objective behind opening an issue and its priority level using feature engineering methods and state-of-the-art text classifiers. To the best of our knowledge, we are the first to fine-tune a Transformer for issue classification. We train and evaluate our models in both project-based and cross-project settings. The latter approach provides a generic prediction model applicable for any unseen software project or projects with little historical data. Our proposed approach can successfully predict the objective and priority level of issue reports with $$82\%$$ (fine-tuned RoBERTa) and $$75\%$$ (Random Forest) accuracy, respectively. Moreover, we conducted human labeling and evaluation on unlabeled issues from six unseen GitHub projects to assess the performance of the cross-project model on new data. The model achieves $$90\%$$ accuracy on the sample set. We measure inter-rater reliability and obtain an average Percent Agreement of $$85.3\%$$ and Randolph’s free-marginal Kappa of 0.71 that translate to a substantial agreement among labelers.

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.