Machine Learning Algorithms Research Articles

Establishing a new-onset diabetes-related metabolism signature for predicting the prognosis and immune landscape in pancreatic cancer

Abstract New-onset diabetes (NOD) is a common condition among patients with pancreatic adenocarcinoma (PAAD) and is related to poor clinical outcomes. The potential impact of NOD on PAAD progression and the tumor microenvironment remains unclear. Here, we revealed that NOD in PAAD was associated with metabolic disorders. Utilizing three machine learning algorithms, a new-onset diabetes-related metabolism signature (NRMS) was established. Validated in three independent cohorts, patients with a high NRMS score exhibited worse prognosis. Moreover, an elevated NRMS score was associated with an immunosuppressive microenvironment and diminished response to immunotherapy. Further experiments demonstrated that ALDH3A1, a key feature in NRMS, was significantly up-regulated in tissues from PAAD patients with NOD and played a crucial role in tumor progression and immune suppression. Our findings highlight the potential of NRMS as a prognostic biomarker and an indicator of immunotherapy response for patients with PAAD.

Raman spectroscopy and SERS for determining the relative age of fossil resins

Abstract This paper presents Raman data from studies on 27 fossil resin samples sourced from different geographical locations and representing various relative ages. Ages ranging from the Late Triassic (T3) to the modern period (Q4) were investigated. The primary method for estimating the age of the resins involved determining the ratio ν(C = C)/σ(CH2), which was 1650–1600 cm−1 and 1450 cm−1 respectively, making it possible to differentiate between fossil resins from T3 and K1 and those from Q and Q4. Surface-enhanced Raman spectroscopy, combined with machine learning algorithms, was used to further clarify the differences between the consecutive K1 and K2 ages.

Non-invasive real-time investigation of colorectal cells tight junctions by Raman microspectroscopy analysis combined with machine learning algorithms for organ-on-chip applications

IntroductionColorectal cancer is the third most common malignancy in developed countries. Diagnosis strongly depends on the pathologist’s expertise and laboratory equipment, and patient survival is influenced by the cancer’s stage at detection. Non-invasive spectroscopic techniques can aid early diagnosis, monitor disease progression, and assess changes in physiological parameters in both heterogeneous samples and advanced platforms like Organ-on-Chip (OoC).MethodsIn this study, Raman microspectroscopy combined with Machine Learning was used to analyse structural and biochemical changes in a Caco-2 cell-based intestinal epithelial model before and after treatment with a calcium chelating agent.ResultsThe Machine Learning (ML) algorithm successfully classified different epithelium damage conditions, achieving an accuracy of 91.9% using only 7 features. Two data-splitting approaches, “sample-based” and “spectra-based,” were also compared. Further, Raman microspectroscopy results were confirmed by TEER measurements and immunofluorescence staining.DiscussionExperimental results demonstrate that this approach, combined with supervised Machine Learning, can investigate dynamic biomolecular changes in real-time with high spatial resolution. This represents a promising non-invasive alternative technique for characterizing cells and biological barriers in organoids and OoC platforms, with potential applications in cytology diagnostics, tumor monitoring, and drug efficacy analysis.

Advancements in the treatment of autoimmune diseases: Integrating artificial intelligence for personalized medicine

The incorporation of artificial intelligence (AI) into medical practice has considerably improved the treatment of autoimmune disorders, opening new avenues for personalized therapy. This study examines advances in AI-driven therapeutic options for autoimmune illnesses, including both current and developing treatments. Traditional therapies for autoimmune illnesses, such as immunosuppressive therapy and biologics, attempt to alleviate symptoms but frequently fall short of offering personalized care. Emerging approaches, such as precision medicine and artificial intelligence, are altering the landscape by harnessing massive volumes of patient data to better customize therapies. AI holds the ability to transform autoimmune disease therapy by enhancing diagnosis, discovering biomarkers, optimizing drug development, and personalized treatment procedures. Real-world applications and case studies are examined to demonstrate how machine learning algorithms have improved treatment tactics for rheumatoid arthritis, systemic lupus erythematosus, and multiple sclerosis. While AI has many advantages, like enhanced diagnosis accuracy and personalized therapy, it also has drawbacks, such as data privacy, the requirement for vast datasets, algorithmic bias, and a lack of explain ability. This study emphasizes the advantages of AI, such as improved patient stratification and predictive modelling, while also discussing its drawbacks, such as ethical problems and the possibility of data exploitation. AI presents intriguing prospects for treating autoimmune diseases, but more research and cooperation are required to overcome current difficulties and completely integrate AI into clinical practice.

Aggregated Housing Price Predictions with No Information About Structural Attributes—Hedonic Models: Linear Regression and a Machine Learning Approach

A number of studies have shown that, in hedonic models, the structural attributes of real property have a greater influence on price than external attributes related to location and the immediate neighbourhood. This makes it necessary to include detailed information about structural attributes when predicting prices using regression models and machine learning algorithms and makes it difficult to study the influence of external attributes. In our study of asking prices on the primary residential market in Warsaw (Poland), we used a methodology we developed to determine price indices aggregated to micro-markets, which we further treated as a dependent variable. The analysed database consisted of 10,135 records relating to 2444 residential developments existing as offers on the market at the end of each quarter in the period 2017–2021. Based on these data, aggregated price level indices were determined for 503 micro-markets in which primary market offers were documented. Using the analysed example, we showed that it is possible to predict the value of aggregated price indices based only on aggregated external attributes—location and neighbourhood. Depending on the model, we obtained an R2 value of 75.8% to 82.9% for the prediction in the set of control observations excluded from building the model.

Spatiotemporal Dynamic Analysis of Eutrophication Status Based on Machine Learning-Based Retrieval Algorithm: Case Study in Liangzi Lake, Hubei, China

The routine monitoring of eutrophication is an important measure for observing the variation in water quality and protecting the ecological health of lakes. However, in situ information reflects eutrophication levels within a limited distance and period. In this study, we retrieved the trophic level index (TLI) based on Landsat 8 remote sensing images and using a machine learning (ML) method in Liangzi Lake in Hubei Province, China. The results showed that random forest (RF) outperformed other ML algorithms in estimating the TLI, evaluated by its higher fitness through the Monte Carlo method (median values of R2, RMSE, and MAE are 0.54, 0.047, and 0.037, respectively). In general, 8% of the areas of Liangzi Lake presented an increasing eutrophication level from 2014 to 2022, and 20.1% of the areas reached a mild eutrophication level in 2022. In addition, we found that temperature and anthropogenic activities may impact the eutrophication conditions of the lake. This work uses remote sensing imagery and a ML method to monitor the dynamics of the lake’s eutrophication status, thereby providing a valuable reference for pollution control measures and enhancing the efficiency of water resource management.

NIMG-83. THE PROGNOSTIC VALUE OF PARAMETRIZED KI-67 MAPS DERIVED FROM CONVENTIONAL MRI AND MACHINE LEARNING IN A LARGE RETROSPECTIVE COHORT OF ADULT-TYPE DIFFUSE GLIOMAS

Abstract BACKGROUND Ki-67 is a well-established proliferation marker relied upon for glioma grade assessment. Advances in machine learning (ML) now allow researchers to non-invasively predict the voxelwise Ki-67 index with reasonable accuracy based on anatomical MRI.The purpose of this study is to evaluate the prognostic ability of these predictions in adult-type diffuse gliomas. METHODS We included adult patients with untreated gliomas of various histopathologic grades who underwent their first resection at a comprehensive cancer center from 1993 to 2018. Anatomic MR images were pre-processed, segmented, normalized, then run through a pre-trained ML algorithm to generate Ki-67 parametric maps. The maximum estimated Ki-67 (maxKi-67) for each patient was documented, and the optimal threshold stratifying the cohort into low- and high-risk patients was computed using 10-fold cross-validation. Kaplan-Meier survival curves were then generated for both risk groups and compared using the log-rank and Breslow tests. MaxKi-67 was correlated with OS using Cox proportional hazard regression analysis adjusted for age and preoperative KPS. RESULTS 1,179 patients (age 52±15; 701 males) were included in this study. 727 tumors were histologic grade IV, while 246 and 206 were histologic grade III and II. The mean maxKi-67 was 15.9± 5.2. A maxKi-67 threshold of 13 was computed from the 10-fold cross-validation to dichotomize the cohort by OS. The low-risk group (maxKi-67<13) had a median OS = 2,435 days, significantly different from the high-risk group (maxKi-67>13) which had median OS = 551 days (p<.001). A maxKi-67 index above this threshold significantly increased the hazard of death: unadjusted HR = 2.53 [95% CI 2.17-2.96] and adjusted HR = 1.55 [1.31-1.83]. Age > 55 and preoperative KPS < 90 were also negative prognosticators in the Cox model: adjusted HR for age = 2.79 [2.36-3.31] and for KPS = 2.08 [1.77-2.46]. CONCLUSION Image-based derivation of maxKi-67 index using ML accurately predicts survival in glioma patients with a cut-off of 13 segregating these patients into 2 risk groups.

Enhancing type 2 diabetes mellitus prediction by integrating metabolomics and tree-based boosting approaches

BackgroundType 2 diabetes mellitus (T2DM) is a global health problem characterized by insulin resistance and hyperglycemia. Early detection and accurate prediction of T2DM is crucial for effective management and prevention. This study explores the integration of machine learning (ML) and explainable artificial intelligence (XAI) approaches based on metabolomics panel data to identify biomarkers and develop predictive models for T2DM.MethodsMetabolomics data from T2DM (n = 31) and healthy controls (n = 34) were analyzed for biomarker discovery (mostly amino acids, fatty acids, and purines) and T2DM prediction. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) regression to enhance the model’s accuracy and interpretability. Advanced three tree-based ML algorithms (KTBoost: Kernel-Tree Boosting; XGBoost: eXtreme Gradient Boosting; NGBoost: Natural Gradient Boosting) were employed to predict T2DM using these biomarkers. The SHapley Additive exPlanations (SHAP) method was used to explain the effects of metabolomics biomarkers on the prediction of the model.ResultsThe study identified multiple metabolites associated with T2DM, where LASSO feature selection highlighted important biomarkers. KTBoost [Accuracy: 0.938; CI: (0.880-0.997), Sensitivity: 0.971; CI: (0.847-0.999), Area under the Curve (AUC): 0.965; CI: (0.937-0.994)] demonstrated its effectiveness in using complex metabolomics data for T2DM prediction and achieved better performance than other models. According to KTBoost’s SHAP, high levels of phenylactate (pla) and taurine metabolites, as well as low concentrations of cysteine, laspartate, and lcysteate, are strongly associated with the presence of T2DM.ConclusionThe integration of metabolomics profiling and XAI offers a promising approach to predicting T2DM. The use of tree-based algorithms, in particular KTBoost, provides a robust framework for analyzing complex datasets and improves the prediction accuracy of T2DM onset. Future research should focus on validating these biomarkers and models in larger, more diverse populations to solidify their clinical utility.

Fingerprinting‐Based Indoor Localization in a 3 × 3 Meter Grid Using OFDM Signals at Sub‐6 GHz

ABSTRACTAccurately determining the indoor location of mobile devices has garnered significant interest due to the complex challenges posed by non‐line‐of‐sight (NLOS) propagation and multipath effects. To address this challenge, this paper proposes a new approach to indoor positioning that utilises channel state information (CSI) and machine learning (ML) techniques to improve accuracy. The proposed method extracts the amplitude and phase differences of the subcarriers from the CSI data to create fingerprints. ML algorithms and network architecture are utilised to train the CSI data from two antennas, in the form of phase and amplitude. Experiments conducted in a standard indoor environment demonstrate the effectiveness of the proposed method.

Research progress on the development of myopia prediction models and their predictive performance

The incidence of myopia is continuously increasing worldwide, especially in China, where the prevention and control of myopia is of great significance. Accurate prediction of myopia and early intervention are of great importance for slowing down the development of myopia and reducing the social burden. This article reviews the research progress of myopia prediction models and their predictive performance in recent years. It includes models based on refractive power and ocular biological characteristics, such as those using parameters such as refractive power, axial length, and percentile curves for prediction; models combined with environmental factors, which explore the relationship between myopia in school-age children and the age of onset, parents' myopia, education, and various living environmental factors; models based on genetic factors, which predict by evaluating parents' myopia status and single nucleotide polymorphisms discovered in genome-wide association studies. In addition, the article also introduces the application of artificial intelligence in myopia prediction, such as machine learning and deep learning algorithms that can predict changes in axial length growth and refractive power. These research advances provide a reference for establishing more accurate myopia prediction models and help achieve precise and personalized myopia prevention and control.

Beyond AI and robotics: the dawn of surgical automation in spine surgery

Artificial intelligence (AI), deep learning (DL), and machine learning (ML) algorithms are revolutionizing spine surgery. Soon, these technologies may allow the integration of automated devices into clinical practice. The roles of such devices are yet to be imagined and then developed, but one could assume that automated surgical devices can assist spine surgeons in a variety of ways, such as contextual guidance, precise screw placements, or intraoperative monitoring. In the not-too-distant future, such devices may be able to perform entire surgeries autonomously. Current literature suggests that advancements toward autonomous robotic surgery may improve surgical approaches and reduce negative clinical variation in spine surgery outcomes. This review aims to examine the current trends, practices, and advancements in surgical automation and provide an overview of the stages of automation of devices currently employed within spine surgery.

A comparative analysis of variants of machine learning and time series models in predicting women’s participation in the labor force

Labor force participation of Egyptian women has been a chronic economic problem in Egypt. Despite the improvement in the human capital front, whether on the education or health indicators, female labor force participation remains persistently low. This study proposes a hybrid machine-learning model that integrates principal component analysis (PCA) for feature extraction with various machine learning and time-series models to predict women’s employment in times of crisis. Various machine learning (ML) algorithms, such as support vector machine (SVM), neural network, K-nearest neighbor (KNN), linear regression, random forest, and AdaBoost, in addition to popular time series algorithms, including autoregressive integrated moving average (ARIMA) and vector autoregressive (VAR) models, have been applied to an actual dataset from the public sector. The manpower dataset considered gender from different regions, ages, and educational levels. The dataset was then trained, tested, and evaluated. For performance validation, forecasting accuracy metrics were constructed using mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), mean absolute percent error (MAPE), R-squared (R2), and cross-validated root mean squared error (CVRMSE). Another Dickey-Fuller test was performed to evaluate and compare the accuracy of the applied models, and the results showed that AdaBoost outperforms the other methods by an accuracy of 100%. Compared to alternative works, our findings demonstrate a comprehensive comparative analysis for predicting women’s participation in different regions during an economic crisis.

High-precision ethanol concentration microsensor with global spectra aided by the multi-layer perceptron

Whispering gallery mode (WGM) resonators can be used for precision measurement thanks to their high sensitivity, small size, and fast response time. Nevertheless, the design of such sensors is usually achieved by selecting a typical single-mode tracking method, which leads to low utilization of a great deal of information in the resonance spectrum and affects the precision. Here, we use the multi-layer perceptron (MLP) deep learning algorithm to train the global spectra and realize the high-precision measurement of ethanol concentration. Firstly, a large number of transmission spectra of different ethanol concentrations are collected and directly used as the original data sets. Secondly, the MLP algorithm is used for training and testing. Finally, the local feature dimension is extracted from the global features of the spectrum for prediction. The results show that the prediction accuracy of the global spectra sensing is 99.81%, which is 13.02% higher than that of extracting 10 local features. In addition, the prediction accuracy of the MLP is compared with four other commonly used machine learning (ML) algorithms, and the results show that the MLP algorithm has the highest prediction accuracy. Therefore, the high-precision ethanol concentration sensor proposed in this paper opens a new way for intelligent optical micro-resonator sensing.

IMMU-03. EXPLORING CANCER-SPECIFIC T CELLS AND ANTIGENS IN GLIOBLASTOMA BASED ON SINGLE-CELL SEQUENCING OF CD8+ TILS

Abstract BACKGROUND The effectiveness of cancer immunotherapy against glioblastoma (GBM) remains limited. This study aims to identify cancer-specific antigens to develop antigen-based cancer immunotherapies for GBM. We investigated candidate tumor antigen-specific T cells in GBM by single-cell RNA sequencing (scRNA-seq) and single-cell TCR sequencing (scTCR-seq), and we explored candidate antigens through genetic analysis of tumor tissues. METHODS Flow cytometry analysis was conducted on fresh tumor digest samples to evaluate tumor-infiltrating lymphocytes (TILs) of GBM. Single-cell RNA and TCR sequencing were performed on CD8+ T cells in TILs. Both data generated by Cell Ranger software were loaded into Seurat at R studio. The dimensional reduction for clustering was performed with uniform manifold approximation and projection (UMAP). Additionally, we performed bulk RNA sequencing (RNA-seq) and whole exome sequencing (WES) on tumor tissues. RESULTS Two out of 20 patients (10%) exhibited abundant TILs in GBM. From these two patients, approximately 20,000 CD8+ T cells in total were analyzed in single-cell analysis. Clustering based on UMAP revealed a predominance of clusters expressing exhaustion markers, with high TCR clonality centered on exhausted T cell (TEX) clusters, like our previous date of lung cancer TILs recognizing tumor antigens. Moreover, within the TEX clusters, the expression of CXCL13, often reported as an antigen-specific marker in recent years, was significantly higher in our data compared to two publicly available datasets. Subsequently, based on RNA-seq and WES, 61 candidate neoantigens were estimated using machine-learning prediction algorithms from NEC Corporation. Furthermore, 5 overexpressed cancer/testis antigens, and 3 bacterial species-derived microbial peptides using Kraken2 were selected as antigen candidates. CONCLUSIONS We identified prospective tumor antigen-specific T cell subsets with high CXCL13 expression from two GBM patients. We plan to identify cancer-specific T cells and antigens from these candidates by T cell activation assay.

IMMU-34. PREDICTION OF TUMOR-ASSOCIATED ANTIGENS USING RADIOGENOMICS

Abstract INTRODUCTION Glioblastoma remains a disease with dismal prognosis. Immunotherapy demonstrated promising results against some refractory cancers. However, comparative results have not been accomplished in glioblastoma therapy due to lack of suitable tumor targets and considerable clonal heterogeneity. OBJECTIVE We propose leveraging radiogenomics using machine learning based on Pre-Op MRI to predict tumor antigen expression profiles displaying specific targets for immunotherapy. METHODS We collected RNAseq data and Pre-Op Axial T2-FLAIR images in 29 patients from IVY Glioblastoma Atlas Project. MRIs were segmented using LIFEx software as follows: 1) tumor + edema (T+E) and 2) whole brain (WB). The intended prediction was for tumor-associated antigens. H2O.ai software was used for prediction using an automated supervised machine learning algorithm. Antigens that achieved >55% area under the curve (i.e., prediction) were reanalyzed adjusting for reproducibility, time of analysis and maximum precision. RESULTS An average of 53.2 T2-FLAIR axial slices per patient were used for 3D reconstruction. We selected the following genes (ERBB2, TP53, IL10RA, CD163, EGFR and MGMT) for downstream analysis given their clinical, immunologic, and prognostic relevance. Regarding tumor-associated antigen prediction for ERBB2, TP53, IL10RA, CD163, EGFR and MGMT predicted 52%, 56%, 66%, 66%, 56% and 0%, on T+E segmentation respectively; The values were 49%, 93%, 87%, 47%, 37% and 0% for the respective WB group. Reanalysis revealed 100% prediction accuracy for the presence of mutant TP53 and IL10RA in both segmentation groups (T+E, WB). CD163 and EGFR prediction was 54% and 61% for the T+E group and 48% for both antigens in the WB group. CONCLUSION We successfully predicted the presence of two out of six antigens (i.e., mutant TP53 and IL10RA) in this patient cohort. Our approach can provide expeditious antigen identification to enable cellular therapy manufacturing for patients with unresectable newly diagnosed tumorsor those which relapsed geographically. Furthermore, this MRI-based methodology might provide a precision medicine approach to places that cannot afford costly testing (i.e., RNAseq).

NCMP-10. EFFECT OF EXTENT OF RESECTION ON GLIOMA SEIZURE OUTCOMES BASED ON MOLECULAR GROUPS

Abstract INTRODUCTION Increased extent of resection (EOR) is a known predictor of seizure freedom in glioma patients. However, it is unknown how tumor genomic alterations affect EOR-induced seizure control. OBJECTIVE To evaluate the genotype-specific impact of EOR on post-operative seizure control METHODS Records of 553 glioma patients who underwent tumor resection from 2012-2023 were reviewed. All patient tumors were analyzed with next generation sequencing and FISH analysis. The unsupervised machine learning algorithm non-negative matrix factorization (NMF) was used to cluster the genomic data. Clinical variables including extent of resection of contrast enhancing tumor and non-enhancing tumor, temozolomide therapy, and radiation therapy were analyzed using Cox Proportional Hazards models. RESULTS NMF clustering of tumor sequencing data revealed four molecular groups: Group 1 (n=56): CDKN2a loss, EGFR copy number gain or mutation, TERT promoter (TERTp) mutation, and MGMT methylated; Group 2 (n=27): IDH mutation, TP53 mutation, and MGMT methylated; Group 3 (n=60): TP53 mutation, TERTp mutation and MGMT methylated; Group 4 (n=244): MGMT unmethylated. Thresholds in the post-operative residual non-enhancing tumor volume (NETV) were identified which marked the most difference in seizure control. For patients in Group 1, the NETV threshold that led to improved epilepsy outcomes was <94.7 cc (p=0.047), whereas the threshold for Group 4 was <89.1 cc (p=0.016). When analysis was conducted for Group 2, the threshold of 10.2 cc failed to reach significance (p=0.1). Interestingly, the threshold analysis for Group 3 found that having a NETV of > 10.1 cc showed improved seizure outcomes (p=0.016). A multivariable cox proportional hazard analysis found that twelve-month seizure freedom was predictive of better OS in group 1 (HR 0.02, p=0.05) and group 4 (HR 0.21, p=1.63e-9). CONCLUSION Our study reveals clinically distinct molecular groups of glioma that display different seizure responses. This data suggests a tumor-specific approach to aggressiveness of extent of resection in order to achieve optimal seizure control.

Advancing user classification models: A comparative analysis of machine learning approaches to enhance faculty password policies at the University of Buraimi

In this paper, we assess the results of experiment with different machine learning algorithms for the data classification on the basis of accuracy, precision, recall and F1-Score metrics. We collected metrics like Accuracy, F1-Score, Precision, and Recall: From the Neural Network model, it produced the highest Accuracy of 0.129526 also highest F1-Score of 0.118785, showing that it has the correct balance of precision and recall ratio that can pick up important patterns from the dataset. Random Forest was not much behind with an accuracy of 0.128119 and highest precision score of 0.118553 knit a great ability for handling relations in large dataset but with slightly lower recall in comparison with Neural Network. This ranked the Decision Tree model at number three with a 0.111792, Accuracy Score while its Recall score showed it can predict true positives better than Support Vector Machine (SVM), although it predicts more of the positives than it actually is a majority of the times. SVM ranked fourth, with accuracy of 0.095465 and F1-Score of 0.067861, the figure showing difficulty in classification of associated classes. Finally, the K-Neighbors model took the 6th place, with the predetermined accuracy of 0.065531 and the unsatisfactory results with the precision and recall indicating the problems of this algorithm in classification. We found out that Neural Networks and Random Forests are the best algorithms for this classification task, while K-Neighbors is far much inferior than the other classifiers.

Queering algorithms: LGBTQ+ content creators’ non-conforming and non-confronting workarounds to digital normativity in China

Popular social media platforms in China, such as Bilibili, Douyin, and Xiaohongshu (REDnote), use machine learning algorithms to curate personalized feeds and enforce content moderation. But the opacity of such algorithms introduces socially consequential and often problematic mechanisms of classification. Given the playful resistance of queer individuals against a normative social environment, the interaction between algorithms and content creators warrants close examination. This study investigates how LGBTQ+ content creators navigate these algorithm-driven platforms, focusing on their negotiation of digital normativity to share non-heterosexual content and foster queer sociality in China. Avoiding the oversimplification of viewing algorithms and those implementing them as monolithic entities, I explore how content creators creatively manage to produce and distribute their work within the constraints of these platforms, and even by playing with these constraints. I collected data from an ethnography on algorithms, interviews, and an open-source investigation into recommendation algorithms, through a scavenge approach. I argue that LGBTQ+ content creators engage with algorithms in ways that are simultaneously non-conforming and non-confrontational. They imagine algorithmic effects, appropriate outcomes, and play with the opacity of machine learning to navigate three aspects of digital normativity in China: heteronormativity, Internet civilization, and commercialization. Empirically, I identified three strategies used to create and expand queer sociality within these constraints: camouflaging, filtering, and verticalizing. Theoretically, I introduce ‘workarounds’ as a framework that transcends the binary perspectives of conformity and resistance in discussions of authoritarian governance, algorithmic control, and user-platform interaction.

Data-driven pressure prediction for self-pressurization liquid hydrogen tank using transfer learning method

Prediction of pressure evolution in liquid hydrogen (LH2) tanks utilizing data-driven technology has gradually garnered significant attention. However, training an accurate data-driven model for pressure prediction in LH2 tanks is a challenge due to inadequate experimental data. To end this, this paper presents a Backpropagation Neural Network model for pressure prediction in LH2 tanks based on transfer learning on pre-trained models, using experimental data and results from thermodynamics models. Pre-trained models are firstly optimized by Bayesian optimization algorithms with large sample datasets from the thermal multi-zone model program and then performed the fine-tuning method with small sample datasets from experiments. The effectiveness of the proposed approach is validated by experimental data and numerical results. Compared with baseline models trained only with experimental data, the results demonstrate better model performance, with a maximum of 10.51% accuracy improvement. The proposed approach demonstrates considerable potential in facilitating the application of advanced machine learning algorithms for LH2 tank pressure prediction, significantly reducing the dependence on experimental data collection.

Predicting Green Water Footprint of Sugarcane Crop Using Multi-Source Data-Based and Hybrid Machine Learning Algorithms in White Nile State, Sudan

Water scarcity and climate change present substantial obstacles for Sudan, resulting in extensive migration. This study seeks to evaluate the effectiveness of machine learning models in forecasting the green water footprint (GWFP) of sugarcane in the context of climate change. By analyzing various input variables such as climatic conditions, agricultural data, and remote sensing metrics, the research investigates their effects on the sugarcane cultivation period from 2001 to 2020. A total of seven models, including random forest (RF), extreme gradient boosting (XGBoost), and support vector regressor (SVR), in addition to hybrid combinations like RF-XGB, RF-SVR, XGB-SVR, and RF-XGB-SVR, were applied across five scenarios (Sc) which includes different combinations of variables used in the study. The most significant mean bias error (MBE) was recorded in RF with Sc3 (remote sensing parameters), at 5.14 m3 ton−1, followed closely by RF-SVR at 5.05 m3 ton−1, while the minimum MBE was 0.03 m3 ton−1 in RF-SVR with Sc1 (all parameters). SVR exhibited the highest R2 values throughout all scenarios. Notably, the R2 values for dual hybrid models surpassed those of triple hybrid models. The highest Nash–Sutcliffe efficiency (NSE) value of 0.98 was noted in Sc2 (climatic parameters) and XGB-SVR, whereas the lowest NSE of 0.09 was linked to SVR in Sc3. The root mean square error (RMSE) varied across different ML models and scenarios, with Sc3 displaying the weakest performance regarding remote sensing parameters (EVI, NDVI, SAVI, and NDWI). Effective precipitation exerted the most considerable influence on GWFP, contributing 81.67%, followed by relative humidity (RH) at 7.5% and Tmax at 5.24%. The study concludes that individual models were as proficient as, or occasionally surpassed, double and triple hybrid models in predicting GWFP for sugarcane. Moreover, remote sensing indices demonstrated minimal positive influence on GWFP prediction, with Sc3 producing the lowest statistical outcomes across all models. Consequently, the study advocates for the use of hybrid models to mitigate the error term in the prediction of sugarcane GWFP.