Terms Of Prediction Research Articles

Prediction of departure flight delays through the use of predictive tools based on machine learning/deep learning algorithms

Abstract The objective of this research is to predict the delays in the departure of scheduled commercial flights through a methodology that uses predictive tools based on machine learning/deep learning (ML/DL), with supervised training in regression, based on the available flight datasets. Since the novel contribution of this work is, first, to make the comparison of the predictions in terms of means and statistical variance of the different ML/DL models implemented and, second, to determine the coefficients of the importance of the features or flight attributes, using ML methods known as permutation importance, it is possible to rank the importance of flight attributes by their influence in determining the delay time and reduce the problem of selecting the most important flight attributes. From the results obtained, it is worth mentioning that the model that presents the best performance is the ensemble or combinatorial method of random forest regressor models, with an acceptable prediction range (measured with the root-mean-square-error).

Short and long term prediction of seizure freedom in drug-resistant focal epilepsy surgery

BackgroundThe selection of candidates for drug-resistant focal epilepsy surgery is essential to achieve the best post-surgical outcomes. Objective: To develop two prediction models for seizure freedom in the short and long-term follow-up and from them to create a risk calculator in order to individualize the selection of candidates for surgery and future therapies in each patients. MethodsA sample of 64 consecutive patients who underwent epilepsy surgery at two Cuban tertiary health institutions between 2012 and 2020 constituted the basis for the prediction models. Two models were obtained through the novel methodology, based on biomarker selection reached by resampling methods, cross-validation and high-accuracy index measured through the area under the receiving operating curve (ROC) procedure. ResultsThe first, to pre-operative model included five predictors: epilepsy type, seizures per month, ictal pattern, interictal EEG topography and normal or abnormal magnetic resonance imaging,. it's precision was 0.77 at one year, and with four years and more 0.63. The second model including variables from the trans-surgical and post-surgical stages: the interictal discharges in the post-surgical EEG, incomplete or complete resection of the epileptogenic zone, the surgical techniques employed and disappearance of the discharge in post-resection electrocorticography; the precision of this model was 0.82 at one year, and with four years and more 0.97. ConclusionsThe introduction of trans-surgical and post-surgical variables increase the prediction of the pre-surgical model. A risk calculator was developed using these prediction models, which could be useful as an accurate tool to improve the prediction in epilepsy surgery.

Mechanism of solar photo-thermal transformation for baffled liquid on energy and mass transfer efficiency in direct absorption anaerobic reactor

Agricultural and animal husbandry waste causes serious environmental pollution. A new coupling model of flow heat transfer and anaerobic fermentation is established in this study to enhance the operation efficiency of anaerobic treatment equipment and guard total environment. The model has a vital ability to cut down the difference value between truthful data and predicted data. This model is used in a push-flow anaerobic reactor to study the effect of the baffle on the reactor temperature field and gas production efficiency. The results show that the increase of the baffle improves the temperature uniformity and gas production rate inside the reactor. The duration of the reactors with 15 baffles is approximately 1800 s less than that without baffle, and the relative time saving rate is as high as 56.5%. The gas production of the reactor with 15 baffles is 3.18% higher than that of the reactor without baffle. As time goes on, the concentrations of propionic acid, acetic acid and hydrogen all increase first and then decrease, but the larger BT has the lower concentration of each substance. This improved model can be used to make more accurate predictions in terms of energy and production.

Advanced Tree-Based Techniques for Predicting Unconfined Compressive Strength of Rock Material Employing Non-Destructive and Petrographic Tests.

The accurate estimation of rock strength is an essential task in almost all rock-based projects, such as tunnelling and excavation. Numerous efforts to create indirect techniques for calculating unconfined compressive strength (UCS) have been attempted. This is often due to the complexity of collecting and completing the abovementioned lab tests. This study applied two advanced machine learning techniques, including the extreme gradient boosting trees and random forest, for predicting the UCS based on non-destructive tests and petrographic studies. Before applying these models, a feature selection was conducted using a Pearson's Chi-Square test. This technique selected the following inputs for the development of the gradient boosting tree (XGBT) and random forest (RF) models: dry density and ultrasonic velocity as non-destructive tests, and mica, quartz, and plagioclase as petrographic results. In addition to XGBT and RF models, some empirical equations and two single decision trees (DTs) were developed to predict UCS values. The results of this study showed that the XGBT model outperforms the RF for UCS prediction in terms of both system accuracy and error. The linear correlation of XGBT was 0.994, and its mean absolute error was 0.113. In addition, the XGBT model outperformed single DTs and empirical equations. The XGBT and RF models also outperformed KNN (R = 0.708), ANN (R = 0.625), and SVM (R = 0.816) models. The findings of this study imply that the XGBT and RF can be employed efficiently for predicting the UCS values.

Steady state population balance modelling of precipitation processes: Nucleation, growth and size-dependent agglomeration

In this work a numerical methodology to solve the steady state Population Balance Equation (PBE) is developed. Three crystallisation mechanisms are included, namely: nucleation, size-independent growth and size-dependent loose agglomeration. The numerical method is based on the discretisation of the crystal size as distributed variable. In order to describe the loose agglomeration, the numerical methodology solves two PBE: one including the nucleation and growth mechanisms and one accounting for the agglomeration process. From the first PBE, liquid phase composition, supersaturation, developed crystal surface and Crystallite Size Distribution (CSD) are obtained. Similarly, the second PBE leads to the Agglomerate Size Distribution (ASD). The study of the size-dependant agglomeration kernel induces an additional numerical difficulty due to the dependency of both PBE and agglomeration kernel on the particle size. An accelerated fixed point algorithm based on the crossed secant method is adapted to overcome the difficulty and accurately solve the agglomeration PBE. The oxalic precipitation of uranium is simulated using this numerical methodology. First, the experimental results of a reference case are compared with the numerical predictions in terms of particle size distribution, mean size, mass fraction and moments. Then, the operating conditions are varied in order to test the algorithm robustness and performances. In all cases, the crossed secant method ensures the size-dependent agglomeration PBE solution and properly predicts the ASD. The developed numerical methodology predicts the mean particle size under the experimental uncertainty in a reasonable computation time and number of iterations.

Stochastic processes in the structure and functioning of soil biodiversity

Ecologists are increasingly recognizing the importance of stochastic processes in generating spatial and temporal variation in biological communities. This variation is very high in soil, which hosts not <¼ of all biodiversity on Earth and is central to how terrestrial ecosystems respond to perturbations. Measurement errors, demographic stochasticity (individual variability in traits such as birth and death rates), and environmental stochasticity (fluctuations in environmental properties) are the three main sources of stochasticity in ecology. Here, we synthesize how these three sources of stochasticity are quantified and incorporated in the study of soil biodiversity, highlighting current limits, possible solutions, and future research needs. We stress the relevance of all these factors to our future understanding of terrestrial ecosystems via plant-soil and soil-climate interactions and feedbacks. In soil, measurement errors are due to the small size, high abundance, and broad distributions of soil organisms, which limit sampling in space and especially over time. We argue that positive autocorrelation is a main characteristic of soil environmental properties, which may have important consequences on the response of soil biota to perturbations. At a local scale, large populations of soil organisms also imply a minor role of demographic stochasticity. Despite demographic stochasticity being a less significant source of variability than environmental stochasticity, we show that demographic stochasticity can be sizeable, but that within soil systems, stochasticity of environmental conditions must be accounted for. Explicit consideration of stochastic processes in soil biodiversity research is essential to our future understanding of the processes that control soil biodiversity. In classical ecology, stochasticity implies probabilistic predictions in terms of population growth, extinction, species coexistence, and community diversity. In soil, stochasticity implies very variable responses to climate change and the soil-climate feedback. Future studies will have to identify the major sources of environmental stochasticity with a particular focus on the interaction between multiple global change factors.

Distinct element modeling of the in‐plane response of a steel‐framed retrofit solution for URM structures

AbstractThis paper presents the results of a numerical study aimed at assessing the effectiveness of an innovative steel modular retrofit system on improving the in‐plane lateral capacity of unreinforced masonry (URM) load‐bearing structures. The investigated retrofit solution consists of modular steel frames fastened to the external surface of masonry piers through chemical anchors. Distinct element models are built and calibrated against experimental data coming from a series of in‐plane quasi‐static tests performed on piers in both bare and retrofitted conditions. Two different masonry typologies are considered herein: one made up with solid clay bricks and lime mortar, assembled in a header bond pattern, and the other with typical Italian hollow clay units and cement‐lime mortar, arranged in a Flemish bond pattern. A simplified microscale approach, combined with a mesh‐refinement strategy, is adopted to decrease computational demand without losing numerical accuracy. All in‐plane failure mechanisms, such as mortar and unit tensile failures, as well as masonry crushing for high compressive stresses, are included in the developed numerical models. Moreover, a methodology to explicitly account for the contribution of the investigated retrofit in the distinct element method (DEM) framework is presented and discussed. The proposed modeling strategy yielded a good prediction in terms of lateral stiffness, lateral strength, hysteretic response, and crack pattern for unreinforced and strengthened specimens. A novel set of metrics is proposed to quantitatively assess the retrofit performance benefits.

Numerical modeling of transient cuttings transport in deviated wellbores

Hole-cleaning optimization is one of the major challenges while drilling long-reach horizontal wells. Accumulation of cuttings can affect the equivalent circulating density (ECD) by decreasing the open flow area and increasing the average density of the drilling fluid. The transient development or removal of cuttings bed due to variation of operating parameters are often ignored in the majority of prior numerical studies. In this paper, it is aimed to develop a transient cutting transport model to predict ECD while drilling. A one-dimensional transient two-phase hydraulic model based on the drift-flux approach is developed to predict the cuttings transport considering cuttings deposition and entrainment. The numerical method employs an upwind scheme to discretize the governing equations. To account for the effect of operating parameters on the cuttings bed, two empirical correlations were employed from the literature. A new empirical correlation is also developed for the prediction of cuttings bed hold-up through regression analysis of the experimental results of a published study. A computer program is developed to solve the governing equations and predict the ECD taking into account the effect of suspended cuttings and cuttings bed along the wellbore trajectory. The numerical model is validated by comparing the predicted return weight of cuttings with measured data in the literature for the cases of no pipe rotation and pipe rotation. The effects of employed empirical correlations on predicted ECD profiles are also examined for a hypothetical well. It is realized that the predicted cuttings bed area of the proposed correlation falls between the predictions of the existing empirical correlations. It is concluded that the most conservative prediction in terms of cutting accumulation in the wellbore cannot necessarily yield a conservative prediction in terms of ECD for early times.

Evaluation of High-Resolution Land Cover Geographical Data for the WRF Model Simulations

Increased computing power has made it possible to run simulations of the Weather Research and Forecasting (WRF) numerical model in high spatial resolution. However, running high-resolution simulations requires a higher-detail mapping of landforms, land use, and land cover. Often, higher-resolution data have limited coverage or availability. This paper presents the feasibility of using CORINE Land Cover (CLC) land use and land cover data and alternative high-resolution global coverage land use/land cover (LULC) data from Copernicus Global Land Service Land Cover Map (CGLS-LC100) V2.0 in high-resolution WRF simulations (100 × 100 m). Global LULC data with a resolution of 100 m are particularly relevant for areas not covered by CLC. This paper presents the method developed by the authors for reclassifying land cover data from CGLS-LC100 to MODIS land use classes with defined parameters in the WRF model and describes the procedure for their implementation into the model. The obtained simulation results of the basic meteorological parameters from the WRF simulation using CLC, CGLS-LC100 and default geographical data from MODIS were compared to observations from 13 meteorological stations in the Warsaw area. The research has indicated noticeable changes in the forecasts of temperature, relative humidity wind speed, and direction after using higher-resolution LULC data. The verification results show a significant difference in weather predictions in terms of CLC and CGLS-LC100 LULC data implementation. Due to the fact that better results were obtained for CLC simulations than for CGLS-LC100, it is suggested that CLC data are first used for simulations in numerical weather prediction models and to use CGLS-LC100 data when the area is outside of CLC coverage.

Door to diuretics time is longer among heart failure patients with atrial fibrillation

Hole-cleaning optimization is one of the major challenges while drilling long-reach horizontal wells. Accumulation of cuttings can affect the equivalent circulating density (ECD) by decreasing the open flow area and increasing the average density of the drilling fluid. The transient development or removal of cuttings bed due to variation of operating parameters are often ignored in the majority of prior numerical studies. In this paper, it is aimed to develop a transient cutting transport model to predict ECD while drilling. A one-dimensional transient two-phase hydraulic model based on the drift-flux approach is developed to predict the cuttings transport considering cuttings deposition and entrainment. The numerical method employs an upwind scheme to discretize the governing equations. To account for the effect of operating parameters on the cuttings bed, two empirical correlations were employed from the literature. A new empirical correlation is also developed for the prediction of cuttings bed hold-up through regression analysis of the experimental results of a published study. A computer program is developed to solve the governing equations and predict the ECD taking into account the effect of suspended cuttings and cuttings bed along the wellbore trajectory. The numerical model is validated by comparing the predicted return weight of cuttings with measured data in the literature for the cases of no pipe rotation and pipe rotation. The effects of employed empirical correlations on predicted ECD profiles are also examined for a hypothetical well. It is realized that the predicted cuttings bed area of the proposed correlation falls between the predictions of the existing empirical correlations. It is concluded that the most conservative prediction in terms of cutting accumulation in the wellbore cannot necessarily yield a conservative prediction in terms of ECD for early times.

MP-453086-1 THE USEFULNESS OF EPS TO PREDICT APPROPRIATE THERAPIES IN PRIMARY PREVENTION ICD PATIENTS

Current guidelines recommend ICD implantation for primary prevention in patients with LVEF ≤35% of both ischemic and nonischemic etiology, regardless of the results of the electrophysiological study. Some studies recently questioned this policy. We aimed to access the value of the EP study in long term prediction of the arrhythmic events in patients with reduced EF (≤35%) who underwent EPS and ICD implantation. We compared the risk for all-cause mortality in patients with negative EPS with and without ICD implantation and compared the risk for appropriate therapy according to the results of the EP study of patients who underwent ICD implantation in our institution. The median follow-up was 84 months. Of 293 patients, who underwent EPS, 159 had negative result (noninducible to VT). These patients were more likely to be female and have nonischemic etiology than patients with positive EPS. Patients did not differ in presence of diabetes, atrial fibrillation or ejection fraction. Kaplan-Meier survival analysis showed that there was no difference in mortality between the 2 groups (44.9% vs 42.0%, p=0.672). This was also true in both ischemic (53.2% vs 49.6%, p=0.554) and nonischemic (27.3% vs 27.3%, p=0.974) patients. Of 212 patients who underwent ICD implantation, 90 had negative EPS. Kaplan-Meier survival analysis showed that these patients were significantly less likely to receive appropriate therapy (5.7% vs 23.9%, p<0.0001) and appropriate shock (0% vs 9.7%, p=0.002), all appropriate shocks happened in the ischemic subgroup (0% vs 10.9%, p=0.004). The time to first appropriate ICD therapy was significantly longer in EPS negative patients (111 vs 92 months, p=0.001). The incidence of inappropriate therapies was similar in both groups (6.7% vs 5.8%, p=0.792). Age, gender, ischemic etiology, coronary disease risk factors, EF and NYHA class were not significantly associated with appropriate ICD therapy. ICD implantation did not reduce mortality in patients with reduced EF and negative EPS. Patients with negative EPS and implanted ICD had much lower incidence of appropriate ICD therapies and no appropriate shocks. EP study is useful to guide decisions about ICD implantation in patients with EF≤35%.

In-silico target prediction by ensemble chemogenomic model based on multi-scale information of chemical structures and protein sequences

Identification and validation of bioactive small-molecule targets is a significant challenge in drug discovery. In recent years, various in-silico approaches have been proposed to expedite time- and resource-consuming experiments for target detection. Herein, we developed several chemogenomic models for target prediction based on multi-scale information of chemical structures and protein sequences. By combining the information of a compound with multiple protein targets together and putting these compound-target pairs into a well-established model, the scores to indicate whether there are interactions between compounds and targets can be derived, and thus a target prediction task can be completed by sorting the outputted scores. To improve the prediction performance, we constructed several chemogenomic models using multi-scale information of chemical structures and protein sequences, and the ensemble model with the best performance was used as our final model. The model was validated by various strategies and external datasets and the promising target prediction capability of the model, i.e., the fraction of known targets identified in the top-k (1 to 10) list of the potential target candidates suggested by the model, was confirmed. Compared with multiple state-of-art target prediction methods, our model showed equivalent or better predictive ability in terms of the top-k predictions. It is expected that our method can be utilized as a powerful computational tool to narrow down the potential targets for experimental testing.Graphical

Determinants of Softwood Lumber Prices in the US Northwest

Abstract Lumber prices can be volatile and hard to predict from month to month yet are important for many sectors of the economy, ranging from forestry and construction. An economic model of lumber prices was developed and applied to data representing multiple supply and demand determinants of lumber. Using a suite of econometric models, monthly lumber prices were related back to variables including construction permits, US reserve bank credit, tariffs with Canada, exchange rates with Canada, and variables representing shocks associated with the COVID-19 pandemic. Preferred models use relatively small amounts of publicly available information, making them more accessible to industry participants who want to make their own price predictions. Such information can help guide decisions about whether to expand or scale back an operation in preparation for likely future price movements.

Improved accuracy and less fault prediction errors via modified sequential minimal optimization algorithm.

The benefits and opportunities offered by cloud computing are among the fastest-growing technologies in the computer industry. Additionally, it addresses the difficulties and issues that make more users more likely to accept and use the technology. The proposed research comprised of machine learning (ML) algorithms is Naïve Bayes (NB), Library Support Vector Machine (LibSVM), Multinomial Logistic Regression (MLR), Sequential Minimal Optimization (SMO), K Nearest Neighbor (KNN), and Random Forest (RF) to compare the classifier gives better results in accuracy and less fault prediction. In this research, the secondary data results (CPU-Mem Mono) give the highest percentage of accuracy and less fault prediction on the NB classifier in terms of 80/20 (77.01%), 70/30 (76.05%), and 5 folds cross-validation (74.88%), and (CPU-Mem Multi) in terms of 80/20 (89.72%), 70/30 (90.28%), and 5 folds cross-validation (92.83%). Furthermore, on (HDD Mono) the SMO classifier gives the highest percentage of accuracy and less fault prediction fault in terms of 80/20 (87.72%), 70/30 (89.41%), and 5 folds cross-validation (88.38%), and (HDD-Multi) in terms of 80/20 (93.64%), 70/30 (90.91%), and 5 folds cross-validation (88.20%). Whereas, primary data results found RF classifier gives the highest percentage of accuracy and less fault prediction in terms of 80/20 (97.14%), 70/30 (96.19%), and 5 folds cross-validation (95.85%) in the primary data results, but the algorithm complexity (0.17 seconds) is not good. In terms of 80/20 (95.71%), 70/30 (95.71%), and 5 folds cross-validation (95.71%), SMO has the second highest accuracy and less fault prediction, but the algorithm complexity is good (0.3 seconds). The difference in accuracy and less fault prediction between RF and SMO is only (.13%), and the difference in time complexity is (14 seconds). We have decided that we will modify SMO. Finally, the Modified Sequential Minimal Optimization (MSMO) Algorithm method has been proposed to get the highest accuracy & less fault prediction errors in terms of 80/20 (96.42%), 70/30 (96.42%), & 5 fold cross validation (96.50%).

Selection for treatment of patients at high risk of fracture by the short versus long term prediction models — data from the Belgian FRISBEE cohort

We analyzed the selection for treatment of patients at high risk of fracture by the Belgian FRISBEE imminent model and the FRAX® tool. We identified in the FRISBEE cohort subjects who sustained an incident MOF (mean age 76.5 ± 6.8 years). We calculated their estimated 10-year risk of fracture using FRAX® before and after adjustment for recency and the 2-year probability of fracture using the FRISBEE model. After 6.8 years of follow-up, we validated 480 incident and 54 imminent MOFs. Of the subjects who had an imminent fracture, 94.0% had a fracture risk estimated above 20% by the FRAX® before correction for recency and 98.1% after adjustment, with a specificity of 20.2% and 5.9%, respectively. The sensitivity and specificity of the FRISBEE model at 2 years were 72.2% and 55.4%, respectively, for a threshold of 10%. For these thresholds, 47.3% of the patients were identified at high risk in both models before the correction, and 17.2% of them had an imminent MOF. The adjustment for recency did not change this selection. Before the correction, 34.2% of patients were selected for treatment by FRAX® only, and 18.8% would have had an imminent MOF. This percentage increased to 47% after the adjustment for recency, but only 6% of those would suffer a MOF within 2 years. In our Belgian FRISBEE cohort, the imminent model was less sensitive but more selective in the selection of subjects in whom an imminent fracture should be prevented, resulting in a lower NNT. The correction for recency in this elderly population further decreased the selectivity of FRAX®. These data should be validated in additional cohorts before using them in everyday practice.

Note on the potential to increase the accuracy of source term calculations for spent nuclear fuel

The accuracy of source term predictions is an important factor which determines the efficiency of interim and final storage of spent nuclear fuel. To minimize the required number of storage containers and to minimize the volume and mass of facilities while maintaining safety margins requires accurate codes to predict the decay heat and the gamma and neutron sources with minimum bias for time points ranging from months to thousands of years. While the relevant nuclear data for the purpose of criticality safety received high attention in the last decades and have been extensively verified with dedicated tests, nuclear data relevant for spent nuclear fuel had smaller priority. Mostly results from a radiochemical analysis of samples taken from commercially irradiated fuel have been used for validation purposes. The comparatively sparse data available from tests which exclusively focus on nuclide vector validation under research conditions means that many factors enter the uncertainty estimate of the measurement-theory comparisons and limits the ability to validate codes to a high accuracy. Firstly, the current status of validation efforts is reviewed. Secondly, fields of improvement are explored. Thirdly, the character of uncertainty distributions in measurement-theory comparisons (C/E) of nuclide vectors is analyzed. Currently there are indications that the C/E data is thick tailed which limits improvement of code validation efforts.

AgeEst: An open access web application for skeletal age-at-death estimation employing machine learning

The present study tests the accuracy of commonly adopted age-at-death estimation markers based on the morphology of the pubic symphysis, iliac auricular surface and cranial sutures on a contemporary documented skeletal collection from Greece (81 males and 59 females). Machine learning techniques are used to assess whether a) machine learning classification models can correctly classify skeletons into their correct age group and b) machine learning regression models can predict the correct age to a satisfactory degree. The constructed models are used in a web application (AgeEst), where users can easily employ them to make predictions for their own skeletal assemblages. The results show that the use of machine learning improves age predictions in terms of bias and inaccuracy compared to the direct application of the original methods. However, there is a strong misclassification of middle-aged individuals, stressing the inherent biases both of the skeletal markers traditionally used in age-at-death prediction and of machine learning methods that, in our case, tend to classify most individuals to one of the two extremes (young or old). We would like to invite colleagues to share with us raw data from other skeletal collections to expand the training dataset to address to some extent issues of age mimicry, while the notebook used for the analysis as well as the code used to construct the web application are openly available to promote the further development of this or similar applications by other scholars.

Deep learning-based assessment of saturated flow boiling heat transfer and two-phase pressure drop for evaporating flow

This work proposes new correlations as well as deep learning based modeling of saturated flow boiling heat transfer and two-phase pressure drops for evaporating flow. First, existing saturation flow boiling heat transfer correlations are compared to experimental database (2,500 data points) of numerous refrigerants for tube diameters ranging from 1 to 7 mm. The newly developed correlation for heat transfer outperformed the existing correlations resulting in an MAE =11.28%. For two-phase pressure drop of the evaporating flow, 1,954 measurement data points of 7 refrigerants were used, for the experimental tube diameter ranged from 0.509 to 14 mm. The newly developed correlation for pressure drop outperformed the existing correlations resulting in an MAE =19.07%. An optimal deep learning model (DL) was developed that further improved the accuracy of the prediction in terms of both heat transfer and pressure drop (R2=0.984 and MAE=4.5 % in terms of heat transfer and R2=0.994 and MAE=7.39 % in terms of two-phase pressure drop). The proposed correlations and deep learning models significantly improve microchannel prediction in terms of heat transfer and two phase pressure drop. Besides, explainable artificial intelligence highlights the dependence and interaction between various features affecting the heat transfer and pressure drop.

Risk Prediction of Digital Human Resource Management Based on Artificial Intelligence

The latest information technologies have greatly accelerated the digitalization progress of Human Resource Management (HRM) and many useful techniques and tools have been developed for that purpose. However, in terms of risk management, effective enough tools and methods are still insufficient. Existing studies generally fail to give a turnkey solution to the operational risks in digital HRM system, and the macro measurement models are not suitable for dealing with the risks in the digital HRM system of each single enterprise. In view of these defects, this paper studied the prediction of risks in digital HRM systems based on Artificial Intelligence (AI). Firstly, the paper outlined the functions of a digital HRM system, defined the risk management mechanism of a HRM system, and built a conceptual model for it. Then, this paper proposed a novel method for predicting the risks in the digital HRM system, which innovatively integrates the digital HRM risk event chains with the risk event graph. After that, the paper elaborated on the structures and building principles of the risk event representation layer, risk event chain module, risk event graph module, and attention fusion module. At last, experimental results verified that the proposed model has obvious advantages in digital HRM risk prediction in terms of both stability and accuracy.

Learning Interactions in Reaction Diffusion Equations by Neural Networks.

Partial differential equations are common models in biology for predicting and explaining complex behaviors. Nevertheless, deriving the equations and estimating the corresponding parameters remains challenging from data. In particular, the fine description of the interactions between species requires care for taking into account various regimes such as saturation effects. We apply a method based on neural networks to discover the underlying PDE systems, which involve fractional terms and may also contain integration terms based on observed data. Our proposed framework, called Frac-PDE-Net, adapts the PDE-Net 2.0 by adding layers that are designed to learn fractional and integration terms. The key technical challenge of this task is the identifiability issue. More precisely, one needs to identify the main terms and combine similar terms among a huge number of candidates in fractional form generated by the neural network scheme due to the division operation. In order to overcome this barrier, we set up certain assumptions according to realistic biological behavior. Additionally, we use an L2-norm based term selection criterion and the sparse regression to obtain a parsimonious model. It turns out that the method of Frac-PDE-Net is capable of recovering the main terms with accurate coefficients, allowing for effective long term prediction. We demonstrate the interest of the method on a biological PDE model proposed to study the pollen tube growth problem.