Relative Performance Research Articles

Hybrid black-box classification for customer churn prediction with segmented interpretability analysis

Customer retention management relies on advanced analytics for decision making. Decision makers in this area require methods that are capable of accurately predicting which customers are likely to churn and that allow to discover drivers of customer churn. As a result, customer churn prediction models are frequently evaluated based on both their predictive performance and their capacity to extract meaningful insights from the models. In this paper, we extend hybrid segmented models for customer churn prediction by incorporating powerful models that can capture non-linearities. To ensure the interpretability of such segmented hybrid models, we introduce a novel model-agnostic approach that extends SHAP. We extensively benchmark the proposed methods on 14 customer churn datasets on their predictive performance. The interpretability aspect of the new model-agnostic approach for interpreting hybrid segmented models is illustrated using a case study. Our contributions to decision making literature are threefold. First, we introduce new hybrid segmented models as powerful tools for decision makers to boost predictive performance. Second, we provide insights in the relative predictive performance by an extensive benchmarking study that compares the new hybrid segmented methods with their base models and existing hybrid models. Third, we propose a model-agnostic tool for segmented hybrid models that provide decision makers with a tool to gain insights for any hybrid segmented model and illustrate it on a case study. Although we focus on customer retention management in this study, this paper is also relevant for decision makers that rely on predictive modeling for other tasks.

Multi-source precipitation estimation using machine learning: Clarification and benchmarking

Multi-source precipitation estimation techniques, including interpolation, bias correction, and merging, are essential for acquiring high-precision precipitation datasets. Machine learning (ML) methods have achieved inspiring success in this field, but current studies utilizing ML often blur the boundaries of interpolation, bias correction, and merging because of the utilization of uniform input–output frameworks. The ambiguity in classification (interpolation, bias correction and merging) results in inconsistent and inappropriate comparative benchmarking among studies. A widespread example is the tendency to select easy-to-beat benchmarks, which can magnify the relative performance of ML methods. We propose to reclarify ML-based precipitation estimation by conducting a series of experiments with designed feature combinations to unveil the essence of ML-based methods. Our findings suggest that with different feature combinations, there are no essential distinctions between models for merging, bias correction and interpolation. Ground measurements, as the reference in supervised learning, play a more vital role than different features in precipitation estimation. This means that the ML-based multi-source merging or bias correction should be treated as gauge data-based interpolation with various auxiliary predictors. Moreover, the improvements of ML estimates are more significant compared to low-standard benchmarks like individual precipitation datasets and simple merging methods, but less preeminent compared to high-standard benchmarks such as a sophisticated interpolation method (e.g., Geographically Weighted Regression). Therefore, we call for positioning such ML applications as interpolation and adhering to high-standard benchmarks whenever possible, which can illustrate the true potential and promote further advancements of ML in geospatial estimation.

The Effect of Performance Measure Comparability and Discriminability on Measure Discontinuance

ABSTRACT This study investigates the effect of performance measure characteristics―comparability and discriminability―on performance measure discontinuance. Relative performance evaluation using comparable information offers a means of filtering out common shocks and capturing the subordinates’ own performance relative to that of peers. In addition, a measure’s ability to discriminate good from poor performance is important in the incentive contract. We show that since subordinates have an incentive to put effort into the comparable and discriminable performance measures, the superior is more likely to retain performance measures with greater comparability and discriminability. Using a large sample of performance evaluation data from 15 state-owned enterprises in South Korea over the period 1985–2021, we find that comparable and discriminable performance measures are less likely to be discontinued. Our study contributes to the performance evaluation literature by identifying two measure characteristics that influence the evolution of performance measures.

The effect of a political crisis on performance of community forests and protected areas in Madagascar

Understanding the effectiveness of conservation interventions during times of political instability is important given how much of the world’s biodiversity is concentrated in politically fragile nations. Here, we investigate the effect of a political crisis on the relative performance of community managed forests versus protected areas in terms of reducing deforestation in Madagascar, a biodiversity hotspot. We use remotely sensed data and statistical matching within an event study design to isolate the effect of the crisis and post-crisis period on performance. Annual rates of deforestation accelerated at the end of the crisis and were higher in community forests than in protected areas. After controlling for differences in location and other confounding variables, we find no difference in performance during the crisis, but community-managed forests performed worse in post-crisis years. These findings suggest that, as a political crisis subsides and deforestation pressures intensify, community-based conservation may be less resilient than state protection.

Who made the noise? Systematic approach for the assessment of neutron imaging scintillators

We propose a method to analyze the characteristics of scintillator screens for neutron imaging applications. Using calculations based on the theory of cascaded linear steps as well as experimental measurements, we compared the characteristics of different lithium- and gadolinium-based scintillator screens. Our results show that, despite their much lower light output, gadolinium-based scintillators outperform lithium-based scintillators in terms of noise characteristics for a variety of imaging setups. However, the relative performance of scintillator screens is highly dependent on the other setup characteristics such as the beam spectrum, field of view, used optical lens and size of the camera sensor. Consequently, the selection of the best scintillator screen – as well as the scintillator characteristics assessment in new developments – requires a systematic consideration of all these elements, as enabled by the framework presented here.

Comparison of Tuberculin Skin Testing and Interferon-γ Release Assays in Predicting Tuberculosis Disease

Elimination of tuberculosis (TB) disease in the US hinges on the ability of tests to detect individual risk of developing disease to inform prevention. The relative performance of 3 available TB tests-the tuberculin skin test (TST) and 2 interferon-γ release assays (IGRAs; QuantiFERON-TB Gold In-Tube [QFT-GIT] and SPOT.TB [TSPOT])-in predicting TB disease development in the US remains unknown. To compare the performance of the TST with the QFT-GIT and TSPOT IGRAs in predicting TB disease in high-risk populations. This prospective diagnostic study included participants at high risk of TB infection (TBI) or progression to TB disease at 10 US sites between 2012 and 2020. Participants of any age who had close contact with a case patient with infectious TB, were born in a country with medium or high TB incidence, had traveled recently to a high-incidence country, were living with HIV infection, or were from a population with a high local prevalence were enrolled from July 12, 2012, through May 5, 2017. Participants were assessed for 2 years after enrollment and through registry matches until the study end date (November 15, 2020). Data analysis was performed in June 2023. At enrollment, participants were concurrently tested with 2 IGRAs (QFT-GIT from Qiagen and TSPOT from Oxford Immunotec) and the TST. Participants were classified as case patients with incident TB disease when diagnosed more than 30 days from enrollment. Estimated positive predictive value (PPV) ratios from generalized estimating equation models were used to compare test performance in predicting incident TB. Incremental changes in PPV were estimated to determine whether predictive performance significantly improved with the addition of a second test. Case patients with prevalent TB were examined in sensitivity analysis. A total of 22 020 eligible participants were included in this study. Their median age was 32 (range, 0-102) years, more than half (51.2%) were male, and the median follow-up was 6.4 (range, 0.2-8.3) years. Most participants (82.0%) were born outside the US, and 9.6% were close contacts. Tuberculosis disease was identified in 129 case patients (0.6%): 42 (0.2%) had incident TB and 87 (0.4%) had prevalent TB. The TSPOT and QFT-GIT assays performed significantly better than the TST (PPV ratio, 1.65 [95% CI, 1.35-2.02] and 1.47 [95% CI, 1.22-1.77], respectively). The incremental gain in PPV, given a positive TST result, was statistically significant for positive QFT-GIT and TSPOT results (1.64 [95% CI, 1.40-1.93] and 1.94 [95% CI, 1.65-2.27], respectively). In this diagnostic study assessing predictive value, IGRAs demonstrated superior performance for predicting incident TB compared with the TST. Interferon-γ release assays provided a statistically significant incremental improvement in PPV when a positive TST result was known. These findings suggest that IGRA performance may enhance decisions to treat TBI and prevent TB.

Relative performance evaluation of four bleeding risk scores in atrial fibrillation patients. What does the new DOAC score provide?

BackgroundRecently, the direct oral anticoagulant (DOAC) score was developed and better predicted major bleeding in DOAC-treated patients with atrial fibrillation (AF) than HASBLED did. Little is known on the new score's performance regarding other bleeding risk in AF. MethodsWe studied 14,672 patients diagnosed with AF between 2014 and 2018. During follow-up, we assessed the performance of DOAC score compared with the HASBLED, ORBIT and SWISS scores at predicting major bleeding in DOACs and non-DOACs users. Discrimination, calibration and decision curve analysis (DCA) were used to assess the risk scorer's performance. ResultsThere were 1484 (10.1%) patients on DOACs, 9730 on vitamin K antagonist (VKA), and 3458 on non-oral anticoagulants. Over a median of 3.5 years of follow-up, 79 major bleedings occurred in the DOAC patients, and 486 in the VKA patients (cumulative incidences = 7.4 and 13.9 per 100 patient-years, respectively).Amongst the DOAC patients, the DOAC score discrimination was moderate (C-statistic = 0.711), but significantly higher than HASBLED (C = 0.640; p = 0.03), ORBIT (C = 0.660; p = 0.04), and SWISS scores (C = 0.637; p = 0.002). The DCA showed higher net benefit using DOAC score compared with the remaining scores.In the VKA patients, DOAC score showed the highest discrimination (c-statistic = 0.709), followed by ORBIT (C = 0.692; p = 0.07), HASBLED and SWISS (C = 0.635 and 0.624, respectively; p < 0.01). All risk scores calibrated well, although HASBLED showed relatively poor calibration. ConclusionsThe new DOAC bleeding risk score is a valid and reasonable predictor of major bleeding over a median of 3.5 years of follow-up. Physicians can be reassured about the applicability of DOAC score for bleeding risk stratification in AF patients. RegistrationURL: https://www.clinicaltrials.gov; Unique identifier: NCT04364516

Estimation of Potato Water Footprint Using Machine Learning Algorithm Models in Arid Regions

AbstractPrecise assessment of water footprint to improve the water consumption and crop yield for irrigated agricultural efficiency is required in order to achieve water management sustainability. Although Penman-Monteith is more successful than other methods and it is the most frequently used technique to calculate water footprint, however, it requires a significant number of meteorological parameters at different spatio-temporal scales, which are sometimes inaccessible in many of the developing countries such as Egypt. Machine learning models are widely used to represent complicated phenomena because of their high performance in the non-linear relations of inputs and outputs. Therefore, the objectives of this research were to (1) develop and compare four machine learning models: support vector regression (SVR), random forest (RF), extreme gradient boost (XGB), and artificial neural network (ANN) over three potato governorates (Al-Gharbia, Al-Dakahlia, and Al-Beheira) in the Nile Delta of Egypt and (2) select the best model in the best combination of climate input variables. The available variables used for this study were maximum temperature (Tmax), minimum temperature (Tmin), average temperature (Tave), wind speed (WS), relative humidity (RH), precipitation (P), vapor pressure deficit (VPD), solar radiation (SR), sown area (SA), and crop coefficient (Kc) to predict the potato blue water footprint (BWF) during 1990–2016. Six scenarios (Sc1–Sc6) of input variables were used to test the weight of each variable in four applied models. The results demonstrated that Sc5 with the XGB and ANN model gave the most promising results to predict BWF in this arid region based on vapor pressure deficit, precipitation, solar radiation, crop coefficient data, followed by Sc1. The created models produced comparatively superior outcomes and can contribute to the decision-making process for water management and development planners.

Simple and General Unitarity Conserving Numerical Real-Time Propagators of the Time-Dependent Schrödinger Equation Based on Magnus Expansion.

Magnus expansion (ME) provides a general way to expand the real-time propagator of a time-dependent Hamiltonian within the exponential such that the unitarity is satisfied at any order. We use this property and explicit integration of Lagrange interpolation formulas for the time-dependent Hamiltonian within each time interval and derive approximations that preserve unitarity for the differential time evolution operators of general time-dependent Hamiltonians. The resulting second-order approximation is the same as using the average of Hamiltonians for two end points of time. We identify three fourth-order approximations involving commutators of Hamiltonians at different times and also derive a sixth-order expression. A test of these approximations along with other available expressions for a two-state time-dependent Hamiltonian with sinusoidal time dependences provides information on the relative performance of these approximations and suggests that the derived expressions can serve as useful numerical tools for time evolution in time-resolved spectroscopy, quantum control, quantum sensing, real-time ab initio quantum dynamics, and open system quantum dynamics.

Study on surface morphology and properties of biological type Ag-IPMC and its application on butterfly soft robot

Ionic Polymer Metal Composite (IPMC) is a new type of artificial muscle material. It has unique advantages such as low driving voltage, good flexibility and rapid response that make it suitable for driving soft robots. However, the high preparation cost limited its widely application. The focus of this paper is to study a low-cost and high-performance Ag-IPMC and explores its application in driving bionic butterfly robots. In addition, based on the flexible characteristic of IPMC, a bionic butterfly soft robot is designed and manufactured, and its related performance is also studied. This manuscript also has provided a reference value for the extensive diversified application of the biological IPMC.

An experimental investigation of tensile residual strength of repaired composite laminates after low velocity impact

Low Velocity Impact (LVI) of Carbon Fibre Reinforced Polymer (CFRP) components are common during the manufacture and in-service life of aerostructures. However, the resulting effect on structural mechanical properties has not been extensively studied when the damage occurs on repaired components, which are already compromised to some degree. This research considers the relative mechanical performance of CFRP parts that are (1) pristine, (2) repaired using the stepped scarf method, and (3) repaired then impact damaged (low velocity impact at 5 J energy). For each case, subsurface damage was visually investigated and detailed measurements taken including impact load–time/energy–time behaviour, and post-impact tensile maximum force, stress and strain (and homogenised Young's modulus) before fracture. Further, three different repair patch sizes were considered. The experiments show that negligible surface impact damage was present in all repair cases but that delamination and debonding occurred for all but large repairs. For small and medium repair patches, 71% and 83% of the static strength of pristine samples were observed, with nearly full recovery for large repairs. After LVI, the strength of small repairs did not decrease whereas that of medium repairs declined to 77%. Interestingly, LVI did not have any impact on load bearing capability of large repairs. Stiffness measurements increased with repair size, even exceeding the value for pristine samples in some cases. The maximum stresses sustained were 67%, 77% and 85% that of a pristine sample for small, medium, and large repairs respectively.

Assessing Degree of Overall Prospect for Merger and Acquisition of Managed Funds: A Relative Performance Perspective

Assessing Degree of Overall Prospect for Merger and Acquisition of Managed Funds: A Relative Performance Perspective

Legal actors' and laypersons' utility judgments of eyewitness lineup procedures and outcomes.

Recent attempts to model the relative performances of eyewitness lineup procedures necessarily include theoretical assumptions about the various costs/benefits, or utilities, of different identification outcomes. We collected data to incorporate empirically derived utilities into such modeling as well as data on various stakeholders' views of lineup procedures as tertiary objectives. This research was exploratory; therefore, we did not have a priori hypotheses. We surveyed judges' (n = 70), prosecutors' (n = 28), police officers' (n = 82), and laypersons' (n = 191) opinions about eyewitness identification procedures and the utilities of outcomes of eyewitness identification procedures. We incorporated the utility judgments into models comparing the desirability of various lineup reforms and compared policy preferences between our samples. All samples frequently mentioned estimator and system variables in open-ended evaluations of lineup procedures, but legal samples mentioned system variables more often than did laypersons. Reflector variables (e.g., confidence) were mentioned less often across the board, as was the scientific basis/standardization of identification policy (especially among laypersons). Utility judgments of various identification outcomes indicated that judges adopt values more closely aligned with normative legal ethics (i.e., the Blackstone ratio), whereas other stakeholders (especially laypersons) depart significantly from those standards. Utility models indicated general agreement among samples in lineup procedure preferences, which varied as a function of culprit-presence base rates. Although legal stakeholders vary in how they value eyewitness identification outcomes, their values imply relatively consistent policy preferences that sometimes depart from scientific recommendations. Nonetheless, all samples expressed support for using scientific research to inform legal policy regarding eyewitness evidence. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Clinical utility of C-reactive protein-to-albumin ratio in the management of patients with inflammatory bowel disease.

C-reactive protein (CRP)-to-albumin ratio (CAR) is a novel score with prognostic value in inflammatory conditions. This study assessed the performance of CAR as an objective marker of disease activity and prediction of subtherapeutic infliximab trough levels in patients with inflammatory bowel disease (IBD). A retrospective study was conducted on three different patient cohorts with IBD: patients who had (i) fecal calprotectin (FC) measurements; (ii) Mayo Endoscopic Scores; and (iii) infliximab trough levels available. The relative performances of CAR, albumin, and CRP were compared in predicting disease activity (based on FC or Mayo Endoscopic Score) and infliximab trough levels. In both the FC (n = 289) and endoscopy (n = 65) cohorts, albumin and CAR correlated with objective disease activity. CAR (area under the curve [AUC] 0.70) was only marginally better at detecting active disease, measured by FC, compared to CRP (AUC 0.68). A CAR >0.15 was able to detect Mayo 3 disease (AUC 0.83, sensitivity 81%, specificity 89%). Albumin (r = 0.38) and CAR (r = -0.42) correlated with infliximab trough levels (n = 204). The optimal CAR for detecting subtherapeutic infliximab trough levels was >0.08 (AUC 0.70, sensitivity 66%, specificity 64%). Both albumin and CAR were independent predictors of subtherapeutic infliximab trough levels but correlated poorly with infliximab trough levels longitudinally in the same patient. CAR was only a modest discriminator of subtherapeutic infliximab levels and offers little more than CRP in detecting active disease. CAR has potential to detect severe Mayo 3 disease and could be calculated in patients admitted with suspected acute severe ulcerative colitis.

Plant-Denoising-Net (PDN): A plant point cloud denoising network based on density gradient field learning

Effective point cloud denoising is critical in 3D plant phenotyping applications, which reduces interference in subsequent algorithms and improves the accuracy of plant phenotypes measurement. Deep learning-based point cloud denoising algorithms have shown excellent denoising performance on simple CAD models. However, these algorithms suffer from issues including over-smoothing or shrinkage and low efficiency when applied on density uneven, incomplete, various types of noise and complex plant point clouds. We proposed a plant point cloud denoising network (PDN) based on point cloud density gradient field learning, which can effectively address the challenges posed by plant point clouds. PDN consists of three main modules: point density feature (PDF) exception module, umbrella operator feature (UOF) computation module, and point density gradient (DG) estimation module. The performance of PDN was evaluated in experiments using point clouds of multiple plant species with noise of different types. Under different levels of Gaussian noise, our method achieved a relative performance improvement of 7.6%-19.3% compared to the state-of-the-art baseline methods, reaching state-of-the-art denoising performance. For noise of different types, the majority of our denoising results outperformed the baseline methods. In addition, our method was 0.5 and 8.6 times faster than the baseline methods when processing point clouds with low and high noise level, respectively. The good robustness, generalization, and computational efficacy of PDN are expected to facilitate the acquisition of high-precision 3D point clouds for various plant species, enhance the versatility of 3D phenotyping methods, improve the accuracy of the measurement of structural phenotypes, and increase the throughput of data processing, therefore facilitate the development of modern breeding research. The source code and the datasets used in this study is available on GitHub at https://github.com/suetme/PDN-plant-denoising-net.

Assessment of the salt tolerance of diverse bread wheat (Triticum aestivum L.) genotypes during the early growth stage under hydroponic culture conditions

ObjectivesSoil salinity affects the growth of crop plants, leading to reduced productivity, and is a major challenge for wheat production worldwide. Various adaptations and mitigation approaches in combination with tolerant wheat genotypes can be useful for the sustainability of crop production in saline environments. However, the development of salt-tolerant wheat genotypes is one of the best and most efficient solutions for obtaining desirable yields. Considering these issues, an investigation was carried out under hydroponic nutrient culture conditions to assess the genetic variability and selection of salt-tolerant wheat genotypes by categorizing inequitable morphophysiological and genetic variability as well as multivariate analysis. MethodsTo meet the objectives of this study, 100 wheat genotypes were tested hydroponically in 0 (control) and 15 dS m−1 salt solutions. ConclusionFor all the wheat genotypes grown under saline conditions, the shoot length (SL), root length (RL), shoot fresh weight (SFW), root fresh weight (RFW), total fresh weight (TFW), shoot dry weight (SDW), root dry weight (RDW), and total dry weight (TDW) decreased significantly. Furthermore, significant variation was observed among the genotypes in terms of their characteristics only under saline conditions. In the case of genetic diversity analysis, a high genotypic coefficient of variation (GCV), phenotypic coefficient of variation (PCV), genetic advance in the percentage of the mean (GAM) and high heritability (h2b) were recorded for all tested wheat genotypes based on the SDW, RDW and TDW. Correlation analysis for both genotypic and phenotypic relationships revealed strong positive correlations for TDW, SDW, TFW and SFW. Principal component analysis (PCA) revealed that TDW, TFW, SDW, and SFW were the most discriminative variables for the wheat genotypes, which was confirmed by discriminant function analysis (DFA). PCA-biplot analysis also revealed significant positive correlations between SDW and SFW and between TDW and TFW. Hierarchical cluster analysis was performed for ten clusters based on the relative performance of the genotypes, where the genotypes were characterized into salt-tolerant, medium-salt-tolerant, medium-salt-susceptible and salt-susceptible groups. Among the genotypes, G11, G25 and G29 under cluster VII were categorized as salt tolerant based on their outstanding performance in terms of characteristics only under saline conditions. D2 analysis proved that the wheat genotypes of this cluster were highly divergent from the other cluster genotypes; as a result, these genotypes might be utilized as parents in the development of salt-tolerant wheat genotypes. The current study concluded that SDW and TDW could be employed as criteria for selecting and defining salt-tolerant genotypes during the early growth stage of wheat.

Predicting operative time for metabolic and bariatric surgery using machine learning models: a retrospective observational study.

Predicting operative time is essential for scheduling surgery and managing the operating room. This study aimed to develop machine learning (ML) models to predict the operative time for metabolic and bariatric surgery (MBS) and to compare each model. The authors used the Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program database between 2016 and 2020 to develop ML models, including linear regression, random forest, support vector machine, gradient-boosted tree, and XGBoost model. Patient characteristics and surgical features were included as variables in the model. The authors used the mean absolute error, root mean square error, and R 2 score to evaluate model performance. The authors identified the 10 most important variables in the best-performing model using the Shapley Additive exPlanations algorithm. In total, 668723 patients were included in the study. The XGBoost model outperformed the other ML models, with the lowest root mean square error and highest R 2 score. Random forest performed better than linear regression. The relative performance of the ML algorithms remained consistent across the models, regardless of the surgery type. The surgery type and surgical approach were the most important features to predict the operative time; specifically, sleeve gastrectomy (vs. Roux-en-Y gastric bypass) and the laparoscopic approach (vs. robotic-assisted approach) were associated with a shorter operative time. The XGBoost model best predicted the operative time for MBS among the ML models examined. Our findings can be useful in managing the operating room scheduling and in developing software tools to predict the operative times of MBS in clinical settings.

Effects of Semi-occluded Vocal Tract Exercises with Different Vibration Sources and Duration on Healthy Adults’ Voice

Objectives: This study aimed to explore the impact of semi-occluded vocal tract exercises (SOVTE) with varied vibration sources and durations on the vocal performance of healthy adults. In addition, we sought to confirm the safety of voiced high-frequency oscillation (VHFO) intervention and provide basic data for performances in relation to straw phonation (SP).Methods: The study included 34 healthy adults with no vocal impairments, comprising 8 males and 26 females. The mean age of the participants was 23.00±2.06 years. To assess vocal characteristics during VHFO and SP, voice samples were recorded at three specific time points: prior to performing SOVTE (M0), after 1 minute of exercise (M1), and after 3 minutes of exercise (M3). Cepstral, perturbation, electroglottographic analyses were conducted on the recorded samples of vowel and sentences. Psychometric evaluation was conducted using questionnaires to gauge the self-reported severity of voice and vocal fatigue.Results: No statistically significant differences were observed in the average number of performances or duration per phonation between the two methods. When compared to baseline, CPPV significantly increased after 1 minute and 3 minutes of performance and Acoustic Psychometric Severity Index of Dysphonia decreased after 3 minutes of performance in VHFO. Both methods exhibited no significant differences in terms of psychometric and electroglottographic measurements.Conclusion: Results objectively confirmed the safety of VHFO intervention. This study is clinically significant as it lays the groundwork for future development of optimal treatment regimens and therapeutic programs for voice disorder patients, using foundational data on VHFO.

PMSIEMDL: Design of a Pattern analysis Model for identification of Student Inclination towards different Educational-fields via Multimodal Deep Learning Fusions

Identification of student inclination towards different educational fields requires integration of deep pattern learning models with temporal data analysis techniques. These techniques are highly context sensitive, and cannot be scaled for analysis of students that have an interest in multiple domains. Moreover, existing deep learning models are highly complex, and showcase moderate performance when used on real-time datasets. To overcome these limitations, this text proposes design of a Pattern analysis Model for identification of Student Inclination towards different Educational-fields via Multimodal Deep Learning fusions. The proposed model initially collects data samples from a large number of students, and segregates them into different classes. These include social data, personal habits data, education data, family related data, performance data and future inspiration data classes. These datasets were combined with a customized psychological questionnaire which was curated by experts in the field of student counselling &amp; psychology. Based on student responses, their entity specific classes were generated, that were separately trained via different Convolutional Neural Network (CNN) Models, which assists in identification of student-performance at individual-class levels. These performances are compared with existing inclination datasets via a fusion of Long-Short-Term Memory (LSTM) &amp; Gated Recurrent Neural Network (GRNN), which assists in identification of correlation between subject-level inclinations &amp; their entity classes. This provides with a probabilistic map of different subjects towards which the student might be inclined, and assists them to select their study streams. The generated map was validated for multiple students, and recommendations were made based on higher probability values, which assisted in identification of student inclination levels. The model was evaluated under large datasets and its performance was compared with various state-of-the-art methods under different scenarios. Based on this comparison, it was observed that proposed model was capable of achieving 8.5% better recommendation accuracy, 4.9% higher prediction precision, 6.5% better recommendation recall &amp; 2.9% better Area Under the Curve (AUC) levels, which makes it highly useful for a wide variety of student inclination use cases

Technical note: Comparing three different methods for allocating river points to coarse-resolution hydrological modelling grid cells

Abstract. The allocation of points in a river network to pixels of a coarse-resolution hydrological modelling grid is a well-known issue, especially for hydrologists who use measurements at gauging stations to calibrate and validate distributed hydrological models. To address this issue, the traditional approach involves examining grid cells surrounding the considered river point and selecting the best candidate, based on distance and upstream drainage area as decision criteria. However, recent studies have suggested that focusing on basin boundaries rather than basin areas could prevent many allocation errors, even though the performance gain is rarely assessed. This paper compares different allocation methods and examines their relative performances. Three methods representing various families of methods have been designed: area-based, topology-based and contour-based methods. These methods are implemented to allocate 2580 river points to a 1 km hydrological modelling grid. These points are distributed along the entire hydrographic network of the French southeastern Mediterranean region, covering upstream drainage areas ranging from 5 to 3000 km2. The results indicate that the differences between the methods can be significant, especially for small upstream catchment areas.