Large Sample Dataset Research Articles

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.

Dynamic prediction and optimization of tunneling parameters with high reliability based on a hybrid intelligent algorithm

In this paper, a hybrid intelligent framework comprising Bayesian optimization (BO), gradient boosting with categorical features (CatBoost) and the nondominated sorting genetic algorithm-III (NSGA-III) was proposed to support multiobjective optimization of shield construction parameters without large sample datasets, improve the shield performance, and ensure reliable and interpretable results. First, with the use of the specific tunneling energy, advancing speed and cutter wear as objective functions, a BO-CatBoost prediction model for shield construction parameters and various objectives was constructed, and the key influencing factors were identified via the SHapley Additive exPlanations (SHAP) method. Then, a BO-CatBoost-NSGA-III model was developed to obtain Pareto solutions under different scenarios involving the adjustment of the key influencing factors. Finally, adopting the Wuhan Metro as the background, the accuracy, stability, and generalizability of the constructed algorithm were verified. The results indicated that (1) the developed BO-CatBoost algorithm is superior to 9 other algorithms. The R2 values of the proposed approach were 0.976 and 0.901–0.976 on the test set. (2) The developed BO-CatBoost-NSGA-III algorithm could be used to obtain Pareto solutions under different scenarios via the adjustment of the key influencing factors with the SHAP method, and the optimal solutions could facilitate improvements in the advancing speed, specific tunneling energy and cutter wear of 3.45 %, 6.09 %, and 0.52 %, respectively, with an overall average reliability of 90.5 %. (3) By comparing various prediction algorithms, optimization schemes of different objectives and geological conditions, the accuracy, stability, and generalizability of the constructed algorithm were verified. The developed BO-CatBoost-NSGA-III framework could enable dynamic adjustment of shield construction parameters for decision-making purposes in the event of conflicting shield construction objectives and exhibits generality.

Graph-Based Unsupervised Feature Selection for Interval-Valued Information System.

Feature selection has become one of the hot research topics in the era of big data. At the same time, as an extension of single-valued data, interval-valued data with its inherent uncertainty tend to be more applicable than single-valued data in some fields for characterizing inaccurate and ambiguous information, such as medical test results and qualified product indicators. However, there are relatively few studies on unsupervised attribute reduction for interval-valued information systems (IVISs), and it remains to be studied how to effectively control the dramatic increase of time cost in feature selection of large sample datasets. For these reasons, we propose a feature selection method for IVISs based on graph theory. Then, the model complexity could be greatly reduced after we utilize the properties of the matrix power series to optimize the calculation of the original model. Our approach can be divided into two steps. The first is feature ranking with the principles of relevance and nonredundancy, and the second is selecting top-ranked attributes when the number of features to keep is fixed as a priori. In this article, experiments are performed on 14 public datasets and the corresponding seven comparative algorithms. The results of the experiments verify that our algorithm is effective and efficient for feature selection in IVISs.

A novel soybean mapping index within the global optimal time window

Efficient soybean mapping is critical for agricultural production and yield prediction. However, current sample-driven soybean mapping methods heavily rely on large representative sample datasets, limiting the interpretability of physical mechanisms. Besides, sample-free methods failed to exploit key features that differentiate soybean from other crops, especially Chlorophyll content. Misclassification errors persist and spatiotemporal generalization remains limited. Therefore, this study develops a novel Soybean Mapping Composite Index (SMCI) within a precise Global Optimal Time Window (GOTW). It integrates unique features of soybean Chlorophyll content, canopy water content, and canopy greenness by coupling three red-edge bands (RE2, RE3, and RE4), one near-infrared band, one shortwave infrared band, and two feature indices (Enhanced Vegetation Index and Green Chlorophyll Vegetation Index). The novel index was applied to soybean mapping at six sites in four major soybean producing countries (China, Argentina, Brazil, and the United States) from 2019 to 2021, using an optimal threshold of 3.25. Within the GOTW, the index responds better to spectral features and improves soybean separability. The average overall accuracy (OA: 91%) and average Kappa coefficient (Kappa: 0.83) for the novel index at all sites outperformed the traditional sample-driven Random Forest (RF) method (OA: 84%, Kappa: 0.70) and the existing sample-free index-based Greenness and Water Content Composite Index (GWCCI) (OA: 81%, Kappa: 0.64). Furthermore, interannual transfer experiments consistently showed high accuracy, demonstrating robust spatiotemporal transferability. The proposed SMCI index meets the need for a lightweight and stable soybean mapping tool and serves as a valuable reference for efficient global crop mapping.

EStreams: An integrated dataset and catalogue of streamflow, hydro-climatic and landscape variables for Europe

Large-sample hydrology datasets have become increasingly available, contributing to significant scientific advances. However, in Europe, only a few such datasets have been published, capturing only a fraction of the wealth of information from national data providers in terms of available spatial density and temporal extent. We present “EStreams”, an extensive dataset of hydro-climatic variables and landscape descriptors and a catalogue of openly available stream records for 17,130 European catchments. Spanning up to 120 years, the dataset includes streamflow indices, catchment-aggregated hydro-climatic signatures and landscape attributes (topography, soils, geology, vegetation and landcover). The catalogue provides detailed descriptions that allow users to directly access streamflow data sources, overcoming challenges related to data redistribution policies, language barriers and varied data portal structures. EStreams also provides Python scripts for data retrieval, aggregation and processing, making it dynamic in contrast to static datasets. This approach enables users to update their data as new records become available. Our goal is to extend current large-sample datasets and further integrate hydro-climatic and landscape data across Europe.

Interpretable machine learning on large samples for supporting runoff estimation in ungauged basins

The distribution of flowmeter data and basin characteristic information exhibits substantial disparities, with most flow observations being recorded at a limited number of well-monitored locations. The perennial challenge of achieving reliable and robust hydrological modeling in ungauged catchments through regionalization has persisted. The increasing availability of large-scale hydrological datasets, coupled with recent advancements in machine learning techniques, offers new opportunities to explore patterns of association between basin attributes and hydrological parameters to enhance streamflow predictions. A novel parameter cross-regional transfer approach based on interpretable machine learning (XGBoost) is proposed to accurately predict runoff processes in ungauged regions by leveraging well-trained models across numerous basins within climate zones. We validate the effectiveness of this framework across 5,764 basins in a large sample dataset (Caravan), employing Nash-Sutcliffe Efficiency (NSE), RMSE and bias to assess performance. And a comparison is made with deep transfer learning based on LSTM and Transformer. Results indicate that the proposed method achieves NSE values exceeding 0.2 for 75 % of the ungauged basins, demonstrating superior performance and more stable accuracy compared to pure deep learning models, owing to its incorporation of physical constraints. Furthermore, the response of parameters to basin attributes within different climatic zones in the large-sample context is elucidated through SHAP values, enriching the understanding of hydrological features through data-driven inverse inference. These findings underscore the capability of interpretable machine learning to leverage hydro-physical regularities extracted from abundant basin features, thereby enhancing the accuracy of runoff predictions in ungauged regions.

Detecting malicious DoH traffic: Leveraging small sample analysis and adversarial networks for detection

In light of the escalating frequency of DNS attacks, it is imperative to bolster user security and privacy through the encryption of DNS queries. However, conventional methods for detecting DNS traffic are no longer effective in identifying encrypted traffic, particularly with the utilization of the DNS-over-HTTPS (DoH) protocol, which employs secure HTTPS for DNS resolution. To confront this challenge, we propose a novel model for detecting malicious DoH traffic, named DoH-TriCGAN, which distinguishes between non-DoH, benign DoH, and malicious DoH traffic. DoH-TriCGAN employs a conditional generative adversarial network comprising three network components, for which we only provide additional information to the generator. We extracted different small sample datasets and large sample dataset from the CIRA-CIC-DoHBrw-2020 dataset, to evaluate the efficiency and effectiveness of the proposed DoH-TriCGAN model, and compared the quality of the generated synthetic data. To establish a benchmark, we utilized the six metrics – accuracy, precision, recall, F1-score, ROC_AUC, and PR_AUC – to assess the performance of our model. The results demonstrate our proposed model outperforms the other five models (RF, XGBoost, BiGRU, Autoencoder, Transformer), showing the best performance particularly in scenarios with limited training samples, while also demonstrating data expansion capabilities by generating high-quality synthetic data to address the issue of insufficient network traffic.

A Novel Approach to Surface Roughness Virtual Sample Generation to Address the Small Sample Size Problem in Ultra-Precision Machining.

Surface roughness is one of the main bases for measuring the surface quality of machined parts. A large amount of training data can effectively improve model prediction accuracy. However, obtaining a large and complete surface roughness sample dataset during the ultra-precision machining process is a challenging task. In this article, a novel virtual sample generation scheme (PSOVSGBLS) for surface roughness is designed to address the small sample problem in ultra-precision machining, which utilizes a particle swarm optimization algorithm combined with a broad learning system to generate virtual samples, enriching the diversity of samples by filling the information gaps between the original small samples. Finally, a set of ultra-precision micro-groove cutting experiments was carried out to verify the feasibility of the proposed virtual sample generation scheme, and the results show that the prediction error of the surface roughness prediction model was significantly reduced after adding virtual samples.

FastBiCmrMLM: a fast and powerful compressed variance component mixed logistic model for big genomic case-control genome-wide association study.

Large sample datasets have been regarded as the primary basis for innovative discoveries and the solution to missing heritability in genome-wide association studies. However, their computational complexity cannot consider all comprehensive effects and all polygenic backgrounds, which reduces the effectiveness of large datasets. To address these challenges, we included all effects and polygenic backgrounds in a mixed logistic model for binary traits and compressed four variance components into two. The compressed model combined three computational algorithms to develop an innovative method, called FastBiCmrMLM, for large data analysis. These algorithms were tailored to sample size, computational speed, and reduced memory requirements. To mine additional genes, linkage disequilibrium markers were replaced by bin-based haplotypes, which are analyzed by FastBiCmrMLM, named FastBiCmrMLM-Hap. Simulation studies highlighted the superiority of FastBiCmrMLM over GMMAT, SAIGE and fastGWA-GLMM in identifying dominant, small α (allele substitution effect), and rare variants. In the UK Biobank-scale dataset, we demonstrated that FastBiCmrMLM could detect variants as small as 0.03% and with α ≈ 0. In re-analyses of seven diseases in the WTCCC datasets, 29 candidate genes, with both functional and TWAS evidence, around 36 variants identified only by the new methods, strongly validated the new methods. These methods offer a new way to decipher the genetic architecture of binary traits and address the challenges outlined above.

On Integrating Time-Series Modeling with Long Short-Term Memory and Bayesian Optimization: A Comparative Analysis for Photovoltaic Power Forecasting

The means of energy generation are rapidly progressing as production shifts from a centralized model to a fully decentralized one that relies on renewable energy sources. Energy generation is intermittent and difficult to control owing to the high variability in the weather parameters. Consequently, accurate forecasting has gained increased significance in ensuring a balance between energy supply and demand with maximum efficiency and sustainability. Despite numerous studies on this issue, large sample datasets and measurements of meteorological variables at plant sites are generally required to obtain a higher prediction accuracy. In practical applications, we often encounter the problem of insufficient sample data, which makes it challenging to accurately forecast energy production with limited data. The Holt–Winters exponential smoothing method is a statistical tool that is frequently employed to forecast periodic series, owing to its low demand for training data and high forecasting accuracy. However, this model has limitations, particularly when handling time-series analysis for long-horizon predictions. To overcome this shortcoming, this study proposes an integrated approach that combines the Holt–Winters exponential smoothing method with long short-term memory and Bayesian optimization to handle long-range dependencies. For illustrative purposes, this new method is applied to forecast rooftop photovoltaic production in a real-world case study, where it is assumed that measurements of meteorological variables (such as solar irradiance and temperature) at the plant site are not available. Through our analysis, we found that by utilizing these methods in combination, we can develop more accurate and reliable forecasting models that can inform decision-making and resource management in this field.

Multi-Scale Analysis of Grain Size in the Component Structures of Sediments Accumulated along the Desert-Loess Transition Zone of the Tengger Desert and Implications for Sources and Aeolian Dust Transportation

Aeolian sediments accumulated along the desert-loess transition zone of the Tengger Desert include heterogeneous textures and complex component structures in their grain-size distributions (GSD). However, the sources of these aeolian sediments have not been resolved due to the lack of large reference GSD sample datasets from adjacent regions that contain various types of sediments; such datasets could be used for fingerprinting based on grain-size properties. This lack of knowledge hinders our understanding of the mechanism of aeolian dust releases in these regions and the effects of forcing of atmospheric circulations on the transportation and accumulation of sediments in this region. In this study, we employed a multi-scale grain-size analysis method, i.e., a combination of the single-sample unmixing (SSU) and the parametric end-member modelling (PEMM) techniques, to resolve the component structures of sediments that had accumulated along the desert-loess transition zone of the Tengger Desert. We have also analyzed the component structures of GSDs of various types of sediments, including mobile and fixed sand dunes, lake sediments, and loess sediments from surrounding regions. Our results demonstrate that the patterns observed in coarser fractions of sediments (i.e., sediments with a mode grain size of >100 μm) from the transition zone match well with the patterns of component structures of several types of sediments from the interior of the Tengger Desert, and the patterns seen in the finer fractions (i.e., fine, medium, and coarse silts with a modal size of <63 μm) were broadly consistent with those of loess sediments from the Qilian Mountains. The deflation/erosion of loess from the Qilian Mountains by wind was the most important mechanism underlying the production of these finer grain-size fractions. The East Asia winter monsoon (EAWM) played a key role in transportation of the aeolian dust from these source regions to the desert-loess transition zone of the desert.

CAMELS-Chem: augmenting CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) with atmospheric and stream water chemistry data

Abstract. Large sample datasets are transforming the catchment sciences, but there are few off-the-shelf stream water chemistry datasets with complementary atmospheric deposition, streamflow, meteorology, and catchment physiographic attributes. The existing CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) dataset includes data on topography, climate, streamflow, land cover, soil, and geology across the continental US. With CAMELS-Chem, we pair these existing attribute data for 516 catchments with atmospheric deposition data from the National Atmospheric Deposition Program and water chemistry and instantaneous discharge data from the US Geological Survey over the period from 1980 through 2018 in a relational database and corresponding dataset. The data include 18 common stream water chemistry constituents: Al, Ca, Cl, dissolved organic carbon, total organic carbon, HCO3, K, Mg, Na, total dissolved N, total organic N, NO3, dissolved oxygen, pH (field and lab), Si, SO4, and water temperature. Annual deposition loads and concentrations include hydrogen, NH4, NO3, total inorganic N, Cl, SO4, Ca, K, Mg, and Na. We demonstrate that CAMELS-Chem water chemistry data are sampled effectively across climates, seasons, and discharges for trend analysis and highlight the coincident sampling of stream constituents for process-based understanding. To motivate their use by the larger scientific community across a variety of disciplines, we show examples of how these publicly available datasets can be applied to trend detection and attribution, biogeochemical process understanding, and new hypothesis generation via data-driven techniques.

Microbial functional pathways based on metatranscriptomic profiling enable effective saliva-based health assessments for precision wellness

It is increasingly recognized that an important step towards improving overall health is to accurately measure biomarkers of health from the molecular activities prevalent in the oral cavity. We present a general methodology for computationally quantifying the activity of microbial functional pathways using metatranscriptomic data. We describe their implementation as a collection of eight oral pathway scores using a large salivary sample dataset (n = 9350), and we evaluate score associations with oropharyngeal disease phenotypes within an unseen independent cohort (n = 14,129). Through this validation, we show that the relevant oral pathway scores are significantly worse in individuals with periodontal disease, acid reflux, and nicotine addiction, compared with controls. Given these associations, we make the case to use these oral pathway scores to provide molecular health insights from simple, non-invasive saliva samples, and as molecular endpoints for actionable interventions to address the associated conditions.

An ensemble learning model for continuous cognition assessment based on resting-state EEG

One critical manifestation of neurological deterioration is the sign of cognitive decline. Causes of cognitive decline include but are not limited to: aging, cerebrovascular disease, Alzheimer’s disease, and trauma. Currently, the primary tool used to examine cognitive decline is scale. However, scale examination has drawbacks such as its clinician subjectivity and inconsistent results. This study attempted to use resting-state EEG to construct a cognitive assessment model that is capable of providing a more scientific and robust evaluation on cognition levels. In this study, 75 healthy subjects, 99 patients with Mild Cognitive Impairment (MCI), and 78 patients with dementia were involved. Their resting-state EEG signals were collected twice, and the recording devices varied. By matching these EEG and traditional scale results, the proposed cognition assessment model was trained based on Adaptive Boosting (AdaBoost) and Support Vector Machines (SVM) methods, mapping subjects’ cognitive levels to a 0–100 test score with a mean error of 4.82 (<5%). This study is the first to establish a continuous evaluation model of cognitive decline on a large sample dataset. Its cross-device usability also suggests universality and robustness of this EEG model, offering a more reliable and affordable way to assess cognitive decline for clinical diagnosis and treatment as well. Furthermore, the interpretability of features involved may further contribute to the early diagnosis and superior treatment evaluation of Alzheimer’s disease.

Opportunities and challenges of computer aided diagnosis in new millennium: A bibliometric analysis from 2000 to 2023.

After entering the new millennium, computer-aided diagnosis (CAD) is rapidly developing as an emerging technology worldwide. Expanding the spectrum of CAD-related diseases is a possible future research trend. Nevertheless, bibliometric studies in this area have not yet been reported. This study aimed to explore the hotspots and frontiers of research on CAD from 2000 to 2023, which may provide a reference for researchers in this field. In this paper, we use bibliometrics to analyze CAD-related literature in the Web of Science database between 2000 and 2023. The scientometric softwares VOSviewer and CiteSpace were used to visually analyze the countries, institutions, authors, journals, references and keywords involved in the literature. Keywords burst analysis were utilized to further explore the current state and development trends of research on CAD. A total of 13,970 publications were included in this study, with a noticeably rising annual publication trend. China and the United States are major contributors to the publication, with the United States being the dominant position in CAD research. The American research institutions, lead by the University of Chicago, are pioneers of CAD. Acharya UR, Zheng B and Chan HP are the most prolific authors. Institute of Electrical and Electronics Engineers Transactions on Medical Imaging focuses on CAD and publishes the most articles. New computer technologies related to CAD are in the forefront of attention. Currently, CAD is used extensively in breast diseases, pulmonary diseases and brain diseases. Expanding the spectrum of CAD-related diseases is a possible future research trend. How to overcome the lack of large sample datasets and establish a universally accepted standard for the evaluation of CAD system performance are urgent issues for CAD development and validation. In conclusion, this paper provides valuable information on the current state of CAD research and future developments.

Does Increasing Ethnic Diversity Reduce Electoral Turnout? The Case of New Zealand 1957–2020

A diversifying population may affect electoral turnout when there is an increasing proportion of minorities with weaker voting habits than those of the majority. New Zealand provides an example to test this possibility, by way of growth in the indigenous Māori population and a high level of recent immigration. A large sample dataset provides better estimates of group turnouts than hitherto available. The authors estimate how much ethnic diversification may have affected turnout in New Zealand between the elections of 1957, 2014, and 2020. By adopting a theory that resists conceptualisation of not voting as deviation from ‘civic duty’, the authors avoid ascribing ‘fault’ to this behaviour.

Prediction of the Transit Time of Coronal Mass Ejections with an Ensemble Machine-learning Method

Coronal mass ejections (CMEs), a kind of violent solar eruptive activity, can exert a significant impact on space weather. When arriving at the Earth, they interact with the geomagnetic field, which can boost the energy supply to the geomagnetic field and may further result in geomagnetic storms, thus having potentially catastrophic effects on human activities. Therefore, accurate forecasting of the transit time of CMEs from the Sun to the Earth is vital for mitigating the relevant losses brought by them. XGBoost, an ensemble model that has better performance in some other fields, is applied to the space weather forecast for the first time. During multiple tests with random data splits, the best mean absolute error (MAE) of ∼5.72 hr was obtained, and in this test, 62% of the test CMEs had absolute arrival time error of less than 5.72 hr. The average MAE over all random tests was ∼10 hr. It indicates that our method has a better predictive potential and baseline. Moreover, we introduce two effective feature importance ranking methods. One is the information gain method, a built-in method of ensemble models. The other is the permutation method. These two methods combine the learning process of the model and its performance to rank the CME features, respectively. Compared with the direct correlation analysis on the sample data set, they can help select the important features that closely match the model. These two methods can assist researchers to process large sample data sets, which often require feature selection in advance.

Open-Source 3D Morphing Software for Facial Plastic Surgery and Facial Landmark Detection Research and Open Access Face Data Set Based on Deep Learning (Artificial Intelligence) Generated Synthetic 3D Models.

Background: The scarcity of 3D facial models presents a significant hurdle for researchers and educators. Gathering such data demands substantial resources. Objective: To introduce an open-source 3D morphing software to generate 3D facial data sets for research and to provide a large sample data set that is based on synthetically generated 3D models. Methods: Software is developed to morph 3D facial models in bulk by altering landmark locations. Twenty synthetic 3D facial models are generated utilizing deep learning tools and 28 landmarks located on each. The measurements of synthetic models are confirmed to be realistic by comparing them with facial statistics. Several facial deformities and types are simulated at various magnitudes on 3D models to generate a large data set. Results: An open-source software and an open-access data set of 980 3D facial models, each with 28 landmark locations, are provided. Since the data set is based on synthetically generated 3D models, no institutional review board approval is required. Conclusion: The 3D morphing software and the large 3D data set are expected to benefit researchers and educators in the field of facial surgery and facial landmark detection.

Can continuous simulation be used as an alternative for flood regionalisation? A large sample example from Chile

Flood frequency analysis lies at the core of hydrology and water engineering as one of the most required estimates for water planning and design of hydraulic structures. For ungauged basins, where no information is available, various flood regionalisation techniques have varying degrees of complexity and resulting performance, depending on the study's goal, the region analysed, and the information available. This study evaluates the use of hydrological models for flood regionalisation in Chile, using 1) A large sample dataset of 101 catchments; 2) the continuous simulation approach with the GR4J model; 3) the leave-one-out strategy for performance testing; and, 4) two regionalisation methods: Nearest Neighbour (NN) and Physical Similarity (PS), together with several alternative objective functions for calibration purposes and regionalisation strategy (in all cases adopting a single criterion, single variable and determinist approach for the parameter’s selection). Our results showed that performance (both in calibration–validation and regionalisation) is highly variable (in terms of reproducing the runoff hydrograph and flood statistics), depending on the catchment’s aridity (e.g., around 66–82% of catchments with NSE above 0 in humid regions but it severely drops to 12–44% of catchments with NSE above 0 when evaluating arid catchments). We also found that flood-specific calibration strategies produce better results for floods but poorer performance in runoff hydrograph reproduction. Finally, we highlight that our regionalisation results were in close agreement with those from one of the currently recommended methods by Chilean engineering for flood regionalisation. This is particularly promising, considering that the continuous simulation approach gives access to the complete time series and not only flood statistics. We end this manuscript by discussing several sources of uncertainty, hoping that these can be accounted for in future studies.

Modeling of magnetorheological dampers based on a dual-flow neural network with efficient channel attention

Magnetorheological dampers (MRDs) are intelligent devices for semi-active control and are widely applied in vibration isolation. A high-fidelity modeling method is necessary to take full advantage of the controllable properties of MRDs. Therefore, a nested long short-term memory (NLSTM)-convolutional neural network-efficient channel attention (NLCE) modeling method based on a dual-flow neural network architecture is proposed herein. It uses the time, current, amplitude, frequency, displacement, and velocity as inputs and the damping force as the output. Extensive sinusoidal excitation experiments were conducted using a materials test system and two datasets (large and small sample numbers) were obtained. Five testing sets with different emphases were obtained from different experimental series. Four evaluation indexes were used for a quantitative comparison. First, after training with the large sample dataset, network ablation and comparison experiments were conducted based on a testing set-1. The mean absolute relative error (MARE) evaluation index decreased by 2.290% relative to that of the NLSTM (baseline), indicating that the NLCE method is optimal for predicting the motion characteristics of MRDs. Furthermore, after training with the small sample dataset, comparison experiments were conducted based on testing set-1 and testing set-2. The MAREs decreased by 3.984% and 0.871% relative to that of the NLSTM (baseline), respectively, indicating that the NLCE is also the best modeling method for small sample dataset. The visualization results from the above experiments verified the abilities of the NLCE modeling method for small sample-adaptation, fighting randomness, and identifying similarities. Finally, based on testing set-3, testing set-4 and testing set-5, the NLCE model trained with small sample datasets has high prediction accuracy in predicting the peak damping force (MAREs = 1.456%, 0.880%, and 1.482%, respectively), indicating a high prediction accuracy in the non-hysteretic region. Combining all of the experimental results shows that the NLCE is an effective method for predicting the motion characteristics of MRDs.