Test Set Research Articles

Random forest predictive modeling of prolonged hospital length of stay in elderly hip fracture patients

BackgroundIn elderly individuals suffering from hip fractures, a prolonged hospital length of stay (PLOS) not only heightens the probability of patient complications but also amplifies mortality risks. Yet, most elderly hip fracture patients present compromised baseline health conditions. Additionally, PLOS leads to increased expenses for patient treatment and care, while also diminishing hospital turnover rates. This, in turn, jeopardizes the prompt allocation of beds for urgent cases.MethodsA retrospective study was carried out from October 2021 to November 2023 on 360 elderly hip fracture patients who underwent surgical treatment at West China Hospital. The 75th percentile of the total patient cohort’s hospital stay duration, which was 12 days, was used to define prolonged hospital length of stay (PLOS). The cohort was divided into training and testing datasets with a 70:30 split. A predictive model was developed using the random forest algorithm, and its performance was validated and compared with the Lasso regression model.ResultsOut of 360 patients, 103 (28.61%) experienced PLOS. A Random Forest classification model was developed using the training dataset, identifying 10 essential variables. The Random Forest model achieved perfect performance in the training set, with an area under the curve (AUC), balanced accuracy, Kappa value, and F1 score of 1.000. In the testing set, the model’s performance was assessed with an AUC of 0.846, balanced accuracy of 0.7294, Kappa value of 0.4325, and F1 score of 0.6061.ConclusionThis study aims to develop a prognostic model for predicting delayed discharge in elderly patients with hip fractures, thereby improving the accuracy of predicting PLOS in this population. By utilizing machine learning models, clinicians can optimize the allocation of medical resources and devise effective rehabilitation strategies for geriatric hip fracture patients. Additionally, this method can potentially improve hospital bed turnover rates, providing latent benefits for the healthcare system.

Risk analysis of noise-induced hearing loss of workers in the automobile manufacturing industries based on back-propagation neural network model: a cross-sectional study in Han Chinese population.

This study aims to predict the risk of noise-induced hearing loss (NIHL) through a back-propagation neural network (BPNN) model. It provides an early, simple and accurate prediction method for NIHL. Population based, a cross sectional study. Han, China. This study selected 3266 Han male workers from three automobile manufacturing industries. Information including personal life habits, occupational health test information and occupational exposure history were collected and predictive factors of NIHL were screened from these workers. BPNN and logistic regression models were constructed using these predictors. The input variables of BPNN model were 20, 16 and 21 important factors screened by univariate, stepwise and lasso-logistic regression. When the BPNN model was applied to the test set, it was found to have a sensitivity (TPR) of 83.33%, a specificity (TNR) of 85.92%, an accuracy (ACC) of 85.51%, a positive predictive value (PPV) of 52.85%, a negative predictive value of 96.46% and area under the receiver operating curve (AUC) is: 0.926 (95% CI: 0.891 to 0.961), which demonstrated the better overall properties than univariate-logistic regression modelling (AUC: 0.715) (95% CI: 0.652 to 0.777). The BPNN model has better predictive performance against NIHL than the stepwise-logistic and lasso-logistic regression model in terms of TPR, TNR, ACC, PPV and NPV (p<0.05); the area under the receiver operating characteristics curve of NIHL is also higher than that of the stepwise and lasso-logistic regression model (p<0.05). It was a relatively important factor in NIHL to find cumulative noise exposure, auditory system symptoms, age, listening to music or watching video with headphones, exposure to high temperature and noise exposure time in the trained BPNN model. The BPNN model was a valuable tool in dealing with the occupational risk prediction problem of NIHL. It can be used to predict the risk of an individual NIHL.

Time-Domain Transfer Learning for Accurate Heavy Metal Concentration Retrieval Using Remote Sensing and TrAdaBoost Algorithm: A Case Study of Daxigou, China

Traditionally, the assessment of heavy metal concentrations using remote sensing technology is sample-intensive, with expensive model development. Using a mining area case study of Daxigou, China, we propose a cross-time-domain transfer learning model to monitor heavy metal pollution using samples collected from different time domains. Specifically, spectral indices derived from Landsat 8 multispectral images, terrain, and other auxiliary data correlative to soil heavy metals were prepared. A cross time-domain sample transfer learning model proposed in the paper based on the TrAdaBoost algorithm was used for the Cu content mapping in the topsoil by selective use of soil samples acquired in 2017 and 2019. We found that the proposed model accurately estimated the concentration of Cu in the topsoil of the mining area in 2019 and performed better than the traditional TrAdaBoost algorithms. The goodness of fit (R2) of the test set increased from 0.55 to 0.66; the relative prediction deviation (RPD) increased from 1.37 to 1.76; and finally, the root-mean-square deviation (RMSE), decreased from 8.33 to 7.24 mg·kg−1. The proposed model is potentially applicable to more accurate and inexpensive monitoring of heavy metals, facilitating remediation-related efforts.

Machine Learning Models for Parkinson Disease: Systematic Review.

With the increasing availability of data, computing resources, and easier-to-use software libraries, machine learning (ML) is increasingly used in disease detection and prediction, including for Parkinson disease (PD). Despite the large number of studies published every year, very few ML systems have been adopted for real-world use. In particular, a lack of external validity may result in poor performance of these systems in clinical practice. Additional methodological issues in ML design and reporting can also hinder clinical adoption, even for applications that would benefit from such data-driven systems. To sample the current ML practices in PD applications, we conducted a systematic review of studies published in 2020 and 2021 that used ML models to diagnose PD or track PD progression. We conducted a systematic literature review in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines in PubMed between January 2020 and April 2021, using the following exact string: "Parkinson's" AND ("ML" OR "prediction" OR "classification" OR "detection" or "artificial intelligence" OR "AI"). The search resulted in 1085 publications. After a search query and review, we found 113 publications that used ML for the classification or regression-based prediction of PD or PD-related symptoms. Only 65.5% (74/113) of studies used a holdout test set to avoid potentially inflated accuracies, and approximately half (25/46, 54%) of the studies without a holdout test set did not state this as a potential concern. Surprisingly, 38.9% (44/113) of studies did not report on how or if models were tuned, and an additional 27.4% (31/113) used ad hoc model tuning, which is generally frowned upon in ML model optimization. Only 15% (17/113) of studies performed direct comparisons of results with other models, severely limiting the interpretation of results. This review highlights the notable limitations of current ML systems and techniques that may contribute to a gap between reported performance in research and the real-life applicability of ML models aiming to detect and predict diseases such as PD.

Searching for the social determinants of health: observations from evidence synthesis publications

BackgroundThe social determinants of health (SDOH) are the focus of an exponentially increasing number of publications, including evidence syntheses. However, there is not an established standard for searching for SDOH literature. This study seeks to identify published evidence syntheses pertaining to the SDOH, analyzing the search strategies used and the studies included within these reviews. The primary objectives are to compare search strategies and create a test set of SDOH publications.MethodsWe searched PubMed, Embase, and Scopus for evidence syntheses that mentioned the SDOH in their research questions and included an SDOH search strategy. Relevant data extracted from each review included databases searched; search terms used for the SDOH; conceptual frameworks referenced; and the citations of primary studies included in the reviews, which were compiled to form a test set of cited papers. The relative recall of the respective search strategies was tested by documenting the total number of MEDLINE results each retrieved and the number of test set papers retrieved.ResultsSixty-four evidence syntheses were identified and included in the analysis, and 2750 cited papers were extracted. Findings indicate few commonalities across search strategies in search terms used, the total number of results retrieved, and the number of test set cited papers retrieved. One hundred and ninety-three unique MeSH terms and 1385 unique keywords and phrases were noted among the various search strategies. The number of total results retrieved by the SDOH search strategies ranged from 21,793 to over 16 million. The percentage of cited papers retrieved by the search strategies ranged from 2.46 to 97.9%. Less than 3% of the cited papers were indexed with the Social Determinants of Health MeSH.ConclusionsThere has been little consistency across evidence syntheses in approaches to searching for SDOH literature. Differences in these strategies could have a significant impact on what literature is retrieved, included in reviews, and, consequently, incorporated into evidence-based practice. By documenting these differences and creating a set of papers relevant to SDOH, this research provides a snapshot of the current challenges in searching for SDOH content and lays the groundwork for the creation of a standardized search approach for SDOH literature.

A multi-view fusion lightweight network for CRSwNPs prediction on CT images.

Accurate preoperative differentiation of the chronic rhinosinusitis (CRS) endotype between eosinophilic CRS (eCRS) and non-eosinophilic CRS (non-eCRS) is an important topic in predicting postoperative outcomes and administering personalized treatment. To this end, we have constructed a sinus CT dataset, which comprises CT scan data and pathological biopsy results from 192 patients of chronic rhinosinusitis with nasal polyps (CRSwNP), treated at the Second Affiliated Hospital of Shantou University Medical College between 2020 and 2022. To differentiate CRSwNP endotype on preoperative CT and improve efficiency at the same time, we developed a multi-view fusion model that contains a mini-architecture with each network of 10 layers by modifying the deep residual neural network. The proposed model is trained on a training set and evaluated on a test set. The multi-view deep learning fusion model achieved the area under the receiver-operating characteristics curve (AUC) of 0.991, accuracy of 0.965 and F1-Score of 0.970 in test set. We compared the performance of the mini-architecture with other lightweight networks on the same Sinus CT dataset. The experimental results demonstrate that the developed ResMini architecture contribute to competitive CRSwNP endotype identification modeling in terms of accuracy and parameter number.

Prediction of Lymph Node Metastasis in T1 Colorectal Cancer Using Artificial Intelligence with Hematoxylin and Eosin-Stained Whole-Slide-Images of Endoscopic and Surgical Resection Specimens

According to the current guidelines, additional surgery is performed for endoscopically resected specimens of early colorectal cancer (CRC) with a high risk of lymph node metastasis (LNM). However, the rate of LNM is 2.1–25.0% in cases treated endoscopically followed by surgery, indicating a high rate of unnecessary surgeries. Therefore, this study aimed to develop an artificial intelligence (AI) model using H&amp;E-stained whole slide images (WSIs) without handcrafted features employing surgically and endoscopically resected specimens to predict LNM in T1 CRC. To validate with an independent cohort, we developed a model with four versions comprising various combinations of training and test sets using H&amp;E-stained WSIs from endoscopically (400 patients) and surgically resected specimens (881 patients): Version 1, Train and Test: surgical specimens; Version 2, Train and Test: endoscopic and surgically resected specimens; Version 3, Train: endoscopic and surgical specimens and Test: surgical specimens; Version 4, Train: endoscopic and surgical specimens and Test: endoscopic specimens. The area under the curve (AUC) of the receiver operating characteristic curve was used to determine the accuracy of the AI model for predicting LNM with a 5-fold cross-validation in the training set. Our AI model with H&amp;E-stained WSIs and without annotations showed good performance power with the validation of an independent cohort in a single center. The AUC of our model was 0.758–0.830 in the training set and 0.781–0.824 in the test set, higher than that of previous AI studies with only WSI. Moreover, the AI model with Version 4, which showed the highest sensitivity (92.9%), reduced unnecessary additional surgery by 14.2% more than using the current guidelines (68.3% vs. 82.5%). This revealed the feasibility of using an AI model with only H&amp;E-stained WSIs to predict LNM in T1 CRC.

Predicting treatment failure in stage III colon cancer patients after radical surgery

PurposeThe aim to assess treatment failure in patients with stage III colon cancer who underwent radical surgery and was analyzed using the nomogram.MethodsClinical factors and survival outcomes for stage III colon cancer patients registered in the SEER database from 2018 to 2019 were analyzed, with patients split into training and testing cohorts (7:3 ratio). A total of 360 patients from the First Affiliated Hospital of Longyan served as an external validation cohort. Independent predictors of treatment failure were identified using logistic regression analyses. The nomograms was evaluated by concordance index (C-index), calibration curves, and the area under the curve (AUC), decision curve analysis (DCA) and clinical impact curves (CIC) assessed the clinical utility of nomograms versus TNM staging.ResultsThe study included 4,115 patients with stage III colon cancer. Multivariate logistic analysis age, tumor site, pT stage, pN stage, chemotherapy, pretreatment CEA levels, number of harvested lymph nodes, perineural invasion and marital status were identified as independent risk factors for treatment failure. The C-indices for the training and testing sets were 0.853 and 0.841. Validation by ROC and calibration curves confirmed the stability and reliability of the model. DCA showed that the net clinical effect of the histogram was superior to that of the TNM staging system, while CIC highlighted the potentially large clinical impact of the model.ConclusionsThe developed Nomogram provides a powerful and accurate tool for clinicians to assess the risk of treatment failure after radical surgery in patients with stage III colon cancer.

Forecasting Next-Time-Step Forex Market Stock Prices Using Neural Networks

Purpose: This study aims to predict the closing price of the EUR/JPY currency pair in the forex market using recurrent neural network (RNN) architectures, namely Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), with the incorporation of Bidirectional layers. Methods: The dataset comprises hourly price data obtained from Yahoo Finance and pre-processed accordingly. The data is divided into training and testing sets, and time series sequences are constructed for input into the models. The RNN, LSTM, and GRU models are trained using the Adam optimization algorithm with the mean squared error (MSE) loss metric. Results: Results indicate that the LSTM model, particularly when coupled with Bidirectional layers, exhibits superior predictive performance compared to the other models, as evidenced by lower MSE values. Conclusions: Therefore, the LSTM model with Bidirectional layers is the most effective in predicting the EUR/JPY currency pair's closing price in the forex market. These findings offer valuable insights for practitioners and researchers involved in financial market prediction and neural network modelling

Calibration of citrus intake assessed by food frequency questionnaires using urinary proline betaine in an observational study setting.

Whether observational study can be employed to establish calibration equations for self-reported dietary intake using food biomarkers is unknown. This study aims to demonstrate the feasibility of obtaining calibration equations based on food biomarkers and 7-day diet records (7DDRs) to correct measurement errors of food frequency questionnaires (FFQ) in an observational study setting. The study population consisted of 669 male and 749 female from the Women and Men's Lifestyle Validation Study. In the training set, the biomarker-predicted intake derived by regressing 7DDR-assessed intake on urinary proline betaine level was regressed on the FFQ-assessed intake to obtain the calibration equations. The regression coefficients were applied to the test set to calculate the calibrated FFQ intake. We examined total citrus as well as individual citrus fruits/beverages. Urinary proline betaine was moderately correlated with orange juice intake [Pearson correlation (r): 0.53 for 7DDR and 0.48 for FFQ] but only weakly correlated with intakes of orange (r: 0.12 for 7DDR and 0.15 for FFQ) and grapefruit (r: 0.14 for 7DDR and 0.09 for FFQ). The FFQ-assessed citrus intake was systematically higher than the 7DDR-assessed intake, and after calibrations the mean calibrated FFQ measurements were almost identical to 7DDR assessments. In the test set, the mean intake levels from 7DDRs, FFQ, and calibrated FFQ were 62.5, 75.3, 63.2 g/d for total citrus, 41.6, 42.5, 41.9 g/d for orange juice, 11.8, 24.3, 12.3 g/d for oranges, and 8.3, 9.3, 8.6 g/d for grapefruit. We observed larger differences between calibrated FFQ and 7DDR assessments at the extreme ends of intake, although on average good agreements were observed for all citrus except grapefruit. Our two-step calibration approach has the potential to be adapted to correct systematic measurement error for other foods/nutrients with established food biomarkers in a cost-effective way.

In Silico Prediction of Skin Sensitization for Compounds via Flexible Evidence Combination Based on Machine Learning and Dempster-Shafer Theory.

Skin sensitization is increasingly becoming a significant concern in the development of drugs and cosmetics due to consumer safety and occupational health problems. In silico methods have emerged as alternatives to traditional in vivo animal testing due to ethical and economic considerations. In this study, machine learning methods were used to build quantitative structure-activity relationship (QSAR) models on five skin sensitization data sets (GPMT, LLNA, DPRA, KeratinoSens, and h-CLAT), achieving effective predictive accuracies (correct classification rates of 0.688-0.764 on test sets). To address the complex mechanisms of human skin sensitization, the Dempster-Shafer theory was applied to merge multiple QSAR models, resulting in an evidence-based integrated decision model. Various evidence combinations and combination rules were explored, with the self-defined Q3 rule showing superior balance. The combination of evidence such as GPMT and KeratinoSens and h-CLAT achieved a correct classification rate (CCR) of 0.880 and coverage of 0.893 while maintaining the competitiveness of other combinations. Additionally, the Shapley additive explanations (SHAP) method was used to interpret important features and substructures related to skin sensitization. A comparative analysis of an external human test set demonstrated the superior performance of the proposed method. Finally, to enhance accessibility, the workflow was implemented into a user-friendly software named HSkinSensDS.

Pine Wilt Disease Segmentation with Deep Metric Learning Species Classification for Early-Stage Disease and Potential False Positive Identification

Pine Wilt Disease poses a significant global threat to forests, necessitating swift detection methods. Conventional approaches are resource-intensive but utilizing deep learning on ortho-mapped images obtained from Unmanned Aerial Vehicles offers cost-effective and scalable solutions. This study presents a novel method for Pine Wilt Disease detection and classification using YOLOv8 for segmenting diseased areas, followed by cropping the diseased regions from the original image and applying Deep Metric Learning for classification. We trained a ResNet50 model using semi-hard triplet loss to obtain embeddings, and subsequently trained a Random Forest classifier tasked with identifying tree species and distinguishing false positives. Segmentation was favored over object detection due to its ability to provide pixel-level information, enabling the flexible extension of subsequent bounding boxes. Deep Metric Learning-based classification after segmentation was chosen for its effectiveness in handling visually similar images. The results indicate a mean Intersection over Union of 83.12% for segmentation, with classification accuracies of 98.7% and 90.7% on the validation and test sets, respectively.

Disagreement concerning atopic dermatitis subtypes between an English prospective cohort (ALSPAC) and linked electronic health records.

Subtypes of atopic dermatitis (AD) have been derived from the Avon Longitudinal Study of Parents and Children (ALSPAC) based on presence and severity of symptoms reported in questionnaires (Severe-Frequent, Moderate-Frequent, Moderate-Declining, Mild-Intermittent, Unaffected/Rare). Good agreement between ALSPAC and linked electronic health records (EHRs) would increase trust in the clinical validity of these subtypes and allow inferring subtypes from EHRs alone, which would enable their study in large primary care databases. 1. Explore if presence and number of AD records in EHRs agrees with AD symptom and severity reports from ALSPAC; 2. Explore if EHRs agree with ALSPAC-derived AD subtypes; 3. Construct models to classify ALSPAC-derived AD subtype using EHRs. We used data from the ALSPAC prospective cohort study from 11 timepoints until age 14 years (1991-2008), linked to local general practice EHRs. We assessed how far ALSPAC questionnaire responses and derived subtypes agreed with AD as established in EHRs using different AD definitions (e.g., diagnosis and/or prescription) and other AD-related records. We classified AD subtypes using EHRs, fitting multinomial logistic regression models tuning hyperparameters and evaluating performance in the testing set (ROC AUC, accuracy, sensitivity, and specificity). 8,828 individuals out of a total 13,898 had both been assigned an AD subtype and had linked EHRs. The number of AD-related codes in EHRs generally increased with severity of AD subtype, however not all with the Severe-Frequent subtypes had AD in EHRs, and many with the Unaffected/Rare subtype did have AD in EHRs. When predicting ALSPAC AD subtype using EHRs, the best tuned model had ROC AUC of 0.65, sensitivity of 0.29 and specificity of 0.83 (both macro averaged); when different sets of predictors were used, individuals with missing EHR coverage excluded, and subtypes combined, sensitivity was not considerably improved. ALSPAC and EHRs disagreed not just on AD subtypes, but also on whether children had AD or not. Researchers should be aware that individuals considered as having AD in one source may not be considered as having AD in another.

A preoperative scoring system for predicting the extraprostatic extension of prostate cancer following radical prostatectomy using magnetic resonance imaging and clinical factors.

We aimed to develop a preoperative prediction model for extraprostatic extension (EPE) in prostate cancer (PCa) patients following radical prostatectomy (RP) using MRI and clinical factors. This retrospective study enrolled 266 consecutive patients who underwent RP for PCa in 2022. These patients were divided into a training set (n = 187) and a test set (n = 79) through random assignment. The evaluated variables included age, prostate-specific antigen (PSA) level, prostate volume, PSA density (PSAD), index tumor length on MRI, Prostate Imaging-Reporting and Data System (PI-RADS) category, and EPE-related MRI features as defined by PI-RADS v2.1. A predictive model was constructed through multivariable logistic regression and subsequently translated into a scoring system. The performance of this scoring system in terms of prediction and calibration was assessed using C statistics and the Hosmer‒Lemeshow test. Among patients in the training and test cohorts, 74 (39.6%) and 25 (31.6%), respectively, exhibited EPE after RP. The formulated scoring system incorporated the following factors: PSAD, index tumor length, bulging prostatic contour, and tumor-capsule interface > 10mm as identified on MRI. This scoring system demonstrated strong prediction performance for EPE in both the training (C statistic, 0.87 [95% confidence interval, 0.86-0.87]) and test cohorts (C statistic, 0.85 [0.83-0.89]). Furthermore, the scoring system exhibited good calibration in both cohorts (P = 0.988 and 0.402, respectively). Our scoring system, built upon MRI features defined by the PI-RADS, offers valuable assistance in assessing the likelihood of EPE after RP.

Influence of VR-Assisted College Dance on College Students' Physical and Mental Health and Comprehensive Quality

With the development and the popularization of sports dance, sports dance teaching has become a required elective course in universities. Sports dance can not only improve students' comprehensive quality, but also affect college students' healthy psychology. The use of VR (Virtual Reality) technology in dance education will definitely develop and promote dance education. This paper studies an effective feature extraction method for the characteristics of dance movements based on VR. The edge features of all video images in each segment are accumulated into one image, and the directional gradient histogram features are extracted from it. The results show that compared with the current robust regression method and cascade regression method, our method has higher positioning accuracy on the pollution test set, and more than 75% of the sample errors in this method are within 0.1. This also verifies the effectiveness of this motion recognition algorithm for dance motion recognition. Dance can effectively resist the psychological barriers of college students and improve their comprehensive quality.

Assessment of abiotic reduction rates of organic compounds by interpretable structural factors and experimental conditions in anoxic water environments

For organic contaminants in lake sediments, aquifers, and anaerobic bioreactors, their reduction is one of the primary transformation paths in these anoxic water environments. A simple model is introduced to predict pseudo-first order rate constants (kobs) for the abiotic reduction of organic compounds featuring diverse reducible functional groups. It utilizes the largest experimental dataset of –log kobs, encompassing 59 organic compounds (278 data points). Unlike available complex quantitative structure–activity relationship (QSAR) methods, the novel approach requires both experimental conditions and structural parameters. In comparison to one of the available general QSAR methods, the new model demonstrates favorable performance. The average absolute deviation (AAD), absolute maximum deviation (ADmax), average absolute relative deviation (AARD%), and R-squared (R2) values of the estimated outputs for 54/5 training/test data sets of the new model are 0.641/1.761, 1.761/1.417, 20.52/83.87, and 0.797/0.949, respectively. On the other hand, the available general comparative QSAR method shows the AAD: 1.311/2.301, ADmax: 3.795/3.732, AARD%: 641.0/821.2, and R2: 0.003/0.447. For the test set, AAD, AARD%, ADmax, and R2 values for the new/comparative models are 0.649/2.403, 62.20/190.5, 1.215/3.732 and 0.974/0.789, respectively. In summary, the new model offers a straightforward approach for the manual calculation of –log kobs, demonstrating excellent goodness-of-fit, reliability, precision, and accuracy.

Machine Learning Approaches for Insurance Pricing: A Case Study of Public Liability Coverage in Morocco

As data-driven technologies continue to advance, the importance of machine learning (ML) techniques in enhancing the accuracy of insurance premium calculations cannot be overstated. That is especially crucial because traditional actuarial methods often fail to incorporate individual risk factors fully. This limitation significantly impacts the insurance industry's capacity to accurately determine premium prices, which in turn affects financial stability and customer satisfaction. This research seeks to assess the efficiency insurance premium calculations through different regression models in ML, including polynomial, decision tree, random forest, and gradient boosting. The study employs rigorous analysis techniques using a comprehensive dataset from a Moroccan vehicle insurance company, including claim history and insurance categories. The dataset is split into training and testing sets to assess the accuracy of the ML models using metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R-squared). Initial findings suggest that ML models greatly surpass traditional actuarial methods, indicating the potential for machine learning to transform premium pricing strategies. That could result in more customized and financially sustainable outcomes within the insurance industry. The findings of this paper are likely to contribute to national insurance premium policy and expand the existing literature on this subject in Morocco.

Flexible Deep Learning-Based State of Health Estimation of Lithium-Ion Batteries with Features Extracted from Partial Charging Curves

The state of health is a crucial state that suggests the capacity of lithium-ion batteries to store and restitute energy at a certain power level, which should be carefully monitored in the battery management system. However, the state of health of batteries is unmeasurable and, currently, it is usually estimated within a specific area of the whole charging data, which is very limited in practical application because of the incomplete and random charging behaviors of users. In this paper, we intend to estimate the state of health of batteries with flexible partial charging curves and normal multi-layer perceptron based on the degradation data of eight 0.74 Ah batteries. To make the estimation more adaptive and flexible, we extract several features from partial charging curves. Analysis of the relationship between extracted features and the state of health shows that the extracted features are useful in estimation. As the length of the partial charging curve increases, the extracted features still function well, and the root mean square error of the test set is lower than 1.5%. Further validation on the other two types of batteries reveals that the proposed method achieves high accuracy even with different sampling and working conditions. The proposed method offers an easy-to-implement way to achieve an accurate estimation of a battery’s state of health.

Subway track foundation settlement deformation prediction based on the BiLSTM-AdaBoost model

Abstract The rapid economic expansion has spurred extensive construction near subway networks, im-pacting the stability of their track foundations. Consequently, it's crucial to monitor and predict settlement in subway track foundations. However, the dynamic deformation patterns often exhibit nonlinearity and non-stationarity, posing challenges for traditional linear regression models. To tackle this, our study integrates the BiLSTM (bi-directional long short-term memory) network with the AdaBoost ensemble learning algorithm. Using settlement data from Shanghai metro monitor-ing points, the model is trained and evaluated employing R2 (coefficient of determination), MAE (mean absolute error), and RMSE (root mean square error). Results show that our proposed model displays superior predictive accuracy compared to the LSTM and the BiLSTM, with an average training set R2 of 0.99, test set R2 of 0.78, average MAE of 0.32mm, and average RMSE of 0.4mm. Consequently, for forecasting subway track foundation deformations, employing our network model ensures highly accurate predictive capabilities.

In Search of the Best Low-Cost Methods for Efficient Screening of Conformers.

Locating the lowest energy conformer is crucial for the accurate computation of equilibrium properties of molecular systems. This paper examines the performance of efficient low-cost methods in terms of the alignment and relative energies of their energy minima against the benchmark revDSD-PBEP86-D4/def2-TZVPP//MP2/cc-pVTZ potential energy surface. The low-cost methods considered include GFN-FF, GFN2-xTB, DFTB3, HF-3c, B97-3c, PBEh-3c, and r2SCAN-3c composite methods against a diverse test set of 20 compounds including alkanes, perfluoroalkyl molecules, peptides, open-shell radicals, and Zn(II) complexes of varying sizes. The "3c" composite methods are generally more accurate, but are at least 2-3 orders of magnitude more expensive than tight-binding methods which have energy minima that align well with the benchmark potential energy surface. The findings of this paper were further exploited to introduce a simple strategy involving Grimme's CENSO energy-sorting algorithm that resulted in up to an order of magnitude reduction in computational time for locating the lowest energy conformer on the revDSD-PBEP86-D4/def2-TZVPP//MP2/cc-pVTZ surface.