Model Identification Research Articles

An Improved Parametric 3D Geological Modeling Framework for Seismic Structure Identification Using Deep Learning in Complex Geological Settings

Identifying geological structures from seismic images is critical in understanding the Earth's evolutionary history, assessing geological disasters, and predicting resource distribution. Researchers have used synthetic seismic data with structure labels generated by numerical simulation to train deep learning models to avoid the extensive effort and resources required to manually label seismic data. However, most existing studies are task-oriented and often simplify or even ignore some common and important geological features during the simulation process, which limits the performance of the trained model on real datasets. To further improve the generality and flexibility of deep learning models for structure identification on datasets with complex structural features, we propose an improved parametric 3D structure modeling framework for numerical simulation of synthetic seismic data that include dip, fold, unconformity, channel, cave, salt body, and composite fault structure. Moreover, we revise some simulation solutions to make the simulated geological structures better resemble the real ones. To the best of our knowledge, our synthetic seismic dataset has the richest structural features to date. The results of comparative experiments on multiple synthetic and field datasets reveal that a deep learning model trained on datasets with limited geological features is prone to learn biased representations, which causes the model to misclassify unseen geological structures with similar seismic responses to the target structure. On the contrary, the same deep learning model trained on the proposed dataset provides significantly reduced false predictions of structures with similar seismic responses and improves the identification accuracy of datasets with complex geological features. Although this study only takes fault and channel identification as case studies, the proposed synthetic seismic dataset can be easily adapted to identify other seismic structures.

Algorithm for predicting financial investor behavior based on biomechanical data and planning recognition

In complex and volatile financial markets, investor behavior is an important factor driving market volatility. Traditional studies have mostly relied on methods such as financial market data, questionnaires and psychological scales, but these methods have limitations such as data lag, subjectivity and difficulty in quantification. In recent years, with the development of biomechanics and neuroscience, researchers have begun to explore the use of biomechanical data to predict the behavioral trends of financial investors, which provides a new perspective and methodology for financial market research. This study aims to predict financing behavior in the stock market by constructing an investor sentiment index and combining it with a planning recognition model. Based on the close relationship between biomechanics and emotions, the biomechanical representations of different emotions of investors and their effects on behavior are dissected. Meanwhile, the partial least squares method is used to construct the investor sentiment index, and a planning identification model consisting of Markov chain, planning identification graph and expected utility function is introduced to predict the stock market trend and financing behavior, which is more accurate than the Markov model based on objective data only. In order to enrich the theoretical system of financial market prediction through this study, it provides more accurate market prediction and investment advice for financial institutions and investors. By introducing biomechanical data and planning recognition model, it provides new ideas and methods for understanding the intrinsic connection between investor sentiment and market volatility, and promotes the development of financing business in the domestic market as well as maintains the health and stability of the domestic stock market.

Discriminating the adulteration of varieties and misrepresentation of vintages of Pu’er tea based on Fourier transform near infrared diffuse reflectance spectroscopy

In the Pu’er tea market, the ubiquity of blending different varieties and the fraudulent representation of vintage years present a persistent challenge. Traditional sensory evaluation and experience are often inadequate for discerning the true variety and vintage of tea, highlighting the need for more sophisticated analytical methods to ensure authenticity and quality. Fourier transform near infrared diffuse reflectance spectroscopy combined with radial basis function neural network (RBFNN) was applied for determination of the varieties and vintages of Pu’er tea. For vintage identification, the accuracy, precision, recall, and F1-score of the RBFNN model for the prediction set were 99.2%, 98.2%, 98.0%, and 98.0%, respectively. For identification of varieties adulteration, the corresponding parameters were 98.9%, 97.2%, 96.7%, and 96.6%, respectively. These results illustrated the feasibility to identify the adulteration of varieties and misrepresentation of vintages of Pu’er tea with near infrared spectra and RBFNN model, proving an efficient alternative for Pu’er tea quality inspection, and offering a robust method for combating the pervasive issues within the market.

Interactions between NAD+ metabolism and immune cell infiltration in ulcerative colitis: subtype identification and development of novel diagnostic models

BackgroundUlcerative colitis (UC) is a chronic inflammatory disease of the colonic mucosa with increasing incidence worldwide. Growing evidence highlights the pivotal role of nicotinamide adenine dinucleotide (NAD+) metabolism in UC pathogenesis, prompting our investigation into the subtype-specific molecular underpinnings and diagnostic potential of NAD+ metabolism-related genes (NMRGs).MethodsTranscriptome data from UC patients and healthy controls were downloaded from the GEO database, specifically GSE75214 and GSE87466. We performed unsupervised clustering based on differentially expressed NAD+ metabolism-related genes (DE-NMRGs) to classify UC cases into distinct subtypes. GSEA and GSVA identified potential biological pathways active within these subtypes, while the CIBERSORT algorithm assessed differential immune cell infiltration. Weighted gene co-expression network analysis (WGCNA) combined with differential gene expression analysis was used to pinpoint specific NMRGs in UC. Robust gene features for subtyping and diagnosis were selected using two machine learning algorithms. Nomograms were constructed and their effectiveness was evaluated using receiver operating characteristic (ROC) curves. Reverse transcription quantitative polymerase chain reaction (RT-qPCR) was conducted to verify gene expression in cell lines.ResultsIn our study, UC patients were classified into two subtypes based on DE-NMRGs expression levels, with Cluster A exhibiting enhanced self-repair capabilities during inflammatory responses and Cluster B showing greater inflammation and tissue damage. Through comprehensive bioinformatics analyses, we identified four key biomarkers (AOX1, NAMPT, NNMT, PTGS2) for UC subtyping, and two (NNMT, PARP9) for its diagnosis. These biomarkers are closely linked to various immune cells within the UC microenvironment, particularly NAMPT and PTGS2, which were strongly associated with neutrophil infiltration. Nomograms developed for subtyping and diagnosis demonstrated high predictive accuracy, achieving area under curve (AUC) values up to 0.989 and 0.997 in the training set and up to 0.998 and 0.988 in validation sets. RT-qPCR validation showed a significant upregulation of NNMT and PARP9 in inflamed versus normal colonic epithelia, underscoring their diagnostic relevance.ConclusionOur study reveals two NAD+ subtypes in UC, identifying four biomarkers for subtyping and two for diagnosis. These findings could suggest potential therapeutic targets and contribute to advancing personalized treatment strategies for UC, potentially improving patient outcomes.

Survey paper on Crop Classification using Convolutional Neural Network

Crop Classification using CNN - A Multi-Model Approach for Crop Classification and Health Assessment Using Convolutional Neural Networks and GPT Integration. The integration of deep learning and natural language processing (NLP) in agriculture has gained significant attention for automating crop classification and disease diagnosis. This survey provides an extensive review of various deep learning models applied in crop classification, plant health assessment, and NLP-based report generation. The study explores the effectiveness of object detection models like YOLO, classification networks such as ResNet and EfficientNet, and NLP models like GPT for generating quality assessments. Additionally, this paper discusses the challenges and future research directions in this domain. The survey serves as a precursor to our research implementation, which integrates CNN-based models for crop identification, health assessment, and GPT-based NLP solutions for automated reporting. Key Words: Deep Learning in Agriculture, Convolutional Neural Networks (CNNs), Plant Disease Detection, Generative AI for Crop Analysis, Precision Agriculture.

Bank Business Model Identification, Evolution and Outcomes: Evidence for South Africa

ABSTRACTThis paper presents the findings of an investigation of the type, evolution and impacts on performance of bank business models in South Africa. We identify the various business models used by South African banks using data on the monthly balance sheets of commercial banks made available by the South African Reserve Bank between 1993 and 2022. We cluster banks into different business models based on the composition of their balance sheets. Based on these clusters, we identify business models oriented to wholesale and retail funding, as well as to universal, investment and interbank activities. Overall, our clustering exercise returns six distinct business models. We observe large differences in terms of business size, performance and risk profiles across the business models. We also analyse the evolution of business models over time. The results suggest that banks exhibit relatively stable business models, but where transition exists it tends to be between certain business models. Increased risk is associated with a higher probability of banks shifting business models.

Non-destructive identification of cashmere and wool fibers based on PLS-DA and LDA using NIR spectroscopy

Manual identification of cashmere and wool fibers is often laborious, subjective, and time-consuming due to their extremely similar features. In order to non-destructively and accurately detect these animal fibers, this study proposes a novel detection method based on machine learning algorithms by near-infrared (NIR) spectroscopy. Building upon the preprocessing of NIR spectroscopy data of cashmere and wool fibers, both partial least-squares discriminant analysis (PLS-DA) and linear discriminant analysis (LDA) classifiers are used to distinguish cashmere and wool fibers. First, four data preprocessing methods are applied: mean normalization (MN), z-score standardization (ZSS), mahalanobis distance (MD), and discrete wavelet transform (DWT). Second, following the preprocessing, PLS-DA is used for feature extraction of the spectral data. Finally, based on the criterion of cumulative contribution rate of 80%, determine the number of principal components (PCs) and use the selected PCs as input for LDA. This study compares three feature extraction methods, principal component analysis (PCA), factor analysis, and sparse principal component analysis (SPCA), and two identification models, k-nearest neighbor (KNN) and decision tree (DT). Experimental results indicate that the proposed PLS-DA-LDA model outperforms the other 11 models, offering a new method for the identification of cashmere and wool fibers using NIR spectroscopy.

Machine Learning-Based In Situ Identification of the High-Dimensional Parameter Space for an Equivalent Circuit Model

Abstract Accurate online identification of equivalent circuit model (ECM) parameters of lithium-ion battery electrochemical impedance spectroscopy (EIS) remains a challenge, particularly with high-dimensional parameter spaces. Here, 11-dimensional adapted Randles ECM (AR-ECM) is reduced to two low-dimensional models, and the EIS frequency ranges for each AR-ECM parameter were determined by distinguishing the frequency bands representing different electrochemical processes. A multi-step parameter in-situ identification methodology was developed to minimize onboard training costs by selecting an optimal training set for machine learning based on the Euclidean distance between the collected and generated EIS data. A Gaussian process regression model was constructed by correlating the AR-ECM parameters and EIS to estimate the AR-ECM parameters. Model performance was validated using 12 cells at different temperatures. Experimental results show that a simulated database covering the main EIS and AR-ECM characteristics can be established, whose scale is on the order of 1e^6, much smaller than the order of 1e^11 resulting from each AR-ECM parameter being varied among 10 values. The estimation errors of the key AR-ECM parameters are approximately 5% at different temperatures. The maximum estimation error of all parameters is as low as 9.03%, 13.96% lower than that based on the complex nonlinear least squares method.

Cooperative learning of Pl@ntNet's Artificial Intelligence algorithm: How does it work and how can we improve it?

Abstract Deep learning models for plant species identification rely on large annotated datasets. The Pl@ntNet system enables global data collection by allowing users to upload and annotate plant observations, leading to noisy labels due to diverse user skills. Achieving consensus is crucial for training, but the vast scale of collected data (number of observations, users and species) makes traditional label aggregation strategies challenging. Existing methods either retain all observations, resulting in noisy training data or selectively keep those with sufficient votes, discarding valuable information. Additionally, as many species are rarely observed, user expertise cannot be evaluated as an inter‐user agreement: otherwise, botanical experts would have a lower weight in the AI training step than the average user. Our proposed label aggregation strategy aims to cooperatively train plant identification AI models. This strategy estimates user expertise as a trust score per user based on their ability to identify plant species from crowdsourced data. The trust score is recursively estimated from correctly identified species given the current estimated labels. This interpretable score exploits botanical experts' knowledge and the heterogeneity of users. Subsequently, our strategy removes unreliable observations but retains those with limited trusted annotations, unlike other approaches. We evaluate Pl@ntNet's strategy on a newly released large subset of the Pl@ntNet database focused on European flora, comprising over 6 M observations and 800 K users. This anonymized dataset of votes and observations is released openly via Lefort, Affouard, et al. (2024). We demonstrate that estimating users' skills based on the diversity of their expertise enhances labelling performance. Our findings emphasize the synergy of human annotation and data filtering in improving AI performance for a refined training dataset. We explore incorporating AI‐based votes alongside human input in the label aggregation. This can further enhance human‐AI interactions to detect unreliable observations (even with few votes).

Small unmanned helicopter system identification based on the weighted least square method and improved grey wolf optimisation algorithm

Abstract Aiming to address the issue of low accuracy in model predictions obtained from fitting frequency domain response curves for small unmanned helicopters during the process of modeling their flight dynamics, this study proposes a system identification algorithm based on the combination of weighted least squares and improved grey wolf optimisation algorithm. The algorithm utilises the weighted least squares method to obtain the initial model structure, optimises the initial model parameters using the improved grey wolf optimisation algorithm, and enhances the local search and global optimisation ability of the grey wolf optimisation algorithm by introducing an improved grey wolf subgrouping rule, nonlinear convergence factor and dynamic cooperative rule. Ultimately, this approach establishes a dynamic model for small, unmanned helicopters. The identified model is validated using flight test data, with findings demonstrating that this method achieves higher accuracy in model identification and better fits to frequency domain response curves, thus providing a more accurate reflection of the flight dynamics of small unmanned helicopters.

Ant Colony Optimization for Accelerated Pathway Identification in Connection Element Method Reservoir Models: A Fast-Track Solution for Large-Scale Simulations

In recent years, reservoir models based on the Connection Element Method (CEM) have gained extensive application in reservoir development. This mesh-free modeling approach effectively captures all flow paths and flow-splitting coefficients between nodes, providing a clear view of flow interactions and accurately identifying primary connectivity pathways between injection and production wells. However, the traditional approach of traversing flow paths and splitting coefficients imposes a significant computational load, particularly when applied to large reservoirs with numerous virtual wells. To enhance simulation efficiency, this paper introduces a novel method leveraging the Ant Colony Optimization (ACO) algorithm to efficiently identify the path with the highest splitting coefficient between well pairs. This approach rapidly calculates and filters the dominant connectivity paths between injection and production wells in CEM models. A comparative analysis shows that, while the ACO algorithm provides limited benefit with a small number of connectivity paths, it significantly outperforms the conventional depth-first search algorithm as the number of experimental wells increases.

A Hybrid Approach for Predictive Control of Oil Well Production Using Dynamic System Identification and Real-Time Parameter Estimation

A significant part of modern natural sciences aims to establish a model-based approach to describe the behavior of physical systems and forecast their dynamics in different scenarios. The successful application of model-based analysis of transport phenomena is driven by several components, such as the consistency of a model and the uncertainty associated with its parameters. The unsatisfactory results of simulations can be caused by an improper choice of numerical methods used, but more importantly, it can be a result of wrong assumptions while establishing the model and a poor choice of closure parameters such as physical properties of fluids. This motivated the development of a hybrid approach that combines model identification directly from the data and subsequent real-time parameter estimation, which eventually minimizes the uncertainty of the developed model. This essentially brings a new model-based approach for an optimal simulation of physical phenomena by incorporating stringent interactions between all the stages of the modeling process. The identification of the governing equation from the data is achieved by a regression technique, while the model refinement is performed using the extended Kalman filter algorithm. The obtained in such a way model is then applied for control-oriented analysis. This paper discusses the deployment of such an integrated approach on a step-to-step basis and demonstrates its application to the problem of a single-phase oil inflow to the producing well.

Identification and validation of a prognostic risk model based on radiosensitivity-related genes in nasopharyngeal carcinoma.

Despite advancements with intensity-modulated radiation therapy (IMRT), about 10 % of nasopharyngeal carcinoma (NPC) patients remain resistant to radiotherapy, leading to recurrence and poor prognosis. This study aims to identify radiosensitivity-related genes in NPC and develop a prognostic model to predict patient outcomes. We analyzed 179 NPC samples from Fujian Cancer Hospital using RNA sequencing. Differentially expressed genes (DEGs) were identified between radiotherapy-sensitive and resistant samples. Machine learning algorithms and Cox regression were used to construct a prognostic risk model, validated in the GSE102349 dataset. Additional analyses included functional pathway, immune infiltration, and drug sensitivity. A risk model based on six genes (LCN8, IGSF1, RIMS2, RBP4, TBX10, ETV4) was developed. Kaplan-Meier analysis showed significantly shorter progression-free survival (PFS) in the high-risk group. The model's AUC values were 0.872, 0.807, and 0.802 for 1-year, 3-year, and 5-year predictions. A nomogram including clinical factors was created, and enrichment analysis linked the high-risk group to radiotherapy resistance mechanisms. This study established a novel radiosensitivity-related prognostic model, offering insights into NPC prognosis and radiotherapy resistance mechanisms.

Estimating Models of Supply and Demand: Instruments and Covariance Restrictions

We consider the identification of empirical models of supply and demand with imperfect competition. We show that a restriction on the covariance between unobserved demand and cost shocks can resolve endogeneity and identify the price parameter. We demonstrate how to employ this approach in estimation, and we compare it to the method of instrumental variables. Our formal results also indicate that weaker covariance restrictions can bound the price parameter. We illustrate the covariance restriction approach with applications to ready-to-eat cereal, cement, and airlines. (JEL C51, D12, L13, L61, L66, L93)

PPEIM: A preference path-based early-stage influence accumulation model for influential nodes identification in locally dense multi-core networks

The identification of influential nodes, a critical problem in the field of complex networks, has been extensively studied. However, previous research has primarily focused on maximizing terminal influence across all network structures indiscriminately, making it challenging to accurately identify influential nodes in specific structures. Moreover, overlooking the influence of the temporal dynamics of propagation significantly diminishes the benefits of identifying influential nodes. Therefore, we propose an influential nodes identification model, Preference Path-based Early-stage Influence Accumulation Model (PPEIM), tailored for typical locally dense multi-core networks. The key idea of PPEIM is to identify more influential nodes in early-stage propagation by aggregating dynamic influence propagation volumes superimposed on multiple paths. Specifically, early-stage influence performance is enhanced by sampling paths, mitigating the risk of dense influential nodes resulting from redundant relationships. Moreover, the K-shell, degree, influence distance and link direction are integrated to define connection strength between nodes to guide path selection. And the concept of influence propagation volume is introduced to accurately simulate the influence residuals and losses during the propagation process. To validate the effectiveness and superiority of PPEIM in locally dense multi-core networks, five sets of simulation experiments are conducted on seven real-world datasets. Experimental results demonstrate that PPEIM outperforms six state-of-the-art methods in overall propagation capability, early-stage influence capability, disintegration capability, node dispersion, and discrimination capability.

Machine learning-driven prediction of nitrate-N adsorption efficiency by Fe-modified biochar: Refined model tuning and identification of crucial features

Machine learning-driven prediction of nitrate-N adsorption efficiency by Fe-modified biochar: Refined model tuning and identification of crucial features

Microsatellite stable gastric cancer can be classified into two molecular subtypes with different immunotherapy response and prognosis based on gene sequencing and computational pathology.

Most gastric cancer (GC) patients exhibit microsatellite stability (MSS), yet comprehensive subtyping for prognostic prediction and clinical treatment decisions for MSS GC is lacking. In this work, RNA-sequencing gene expression data and clinical information of MSS GC patients were obtained from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO) databases. We employed several machine learning methods to develop and validate a signature based on immune-related genes (IRGs) for subtyping MSS GC patients. Moreover, two deep learning models based on the Vision Transformer (ViT) architecture were developed to predict GC tumor tiles and identify MSS GC subtypes from digital pathology slides. Microsatellite status was evaluated by immunohistochemistry, and prognostic data as well as H&E whole slide images were collected from 105 MSS GC patients to serve as an independent validation cohort. A signature comprising five IRGs was established and validated, stratifying MSS GC patients into high-risk (MSS-HR) and low-risk (MSS-LR) groups. This signature demonstrated consistent performance, with areas under the receiver operating characteristic (ROC) curve (AUC) of 0.65, 0.70, and 0.70 at 1, 3, and 5 years in the TCGA cohort, and 0.70, 0.60, and 0.62 in the GEO cohort, respectively. The MSS-HR subtype exhibited higher levels of tumor immune dysfunction and exclusion, suggesting a greater potential for immune escape compared to the MSS-LR subtype. Moreover, the MSS-HR/LR subtypes showed differential sensitivities to various therapeutic drugs. Leveraging morphological differences, the tumor recognition segmentation model (TRSM) achieved an impressive AUC of 0.97, while the MSS-HR/LR identification model (MSSIM) effectively classified MSS-HR/LR subtypes with an AUC of 0.94. Both models demonstrated promising results in classifying MSS GC patients in the external validation cohort, highlighting the strong ability to accurately differentiate between MSS GC subtypes. The IRGs-related MSS-HR/LR subtypes had potential in enhancing outcome prediction accuracy and guide treatment strategies. This research may optimize precision treatment and improve the prognosis for MSS GC patients.

Multi-strategy hybrid dung beetle optimization algorithm for parameter identification in photovoltaic systems

Abstract The dung beetle optimization algorithm (DBO) is a novel meta-heuristic algorithm inspired by the behaviors of dung beetle populations, including rolling, dancing, foraging, breeding and stealing. As so far, the DBO algorithm has demonstrated success in addressing a wide range of complex engineering optimization problems. However, like many other meta-heuristic algorithms, it is also prone to certain limitations, such as slow convergence rates and the tendency to become trapped in local optima during the later stages of optimization. To overcome these limitations, this paper proposes a multi-strategy hybrid dung beetle optimization algorithm (MSDBO), which introduces the tangent flight strategy, golden sine search strategy, adaptive t-distribution sparrow perturbation strategy, and vertical crossover mutation strategy. To comprehensively evaluate the performance of MSDBO, simulations are conducted on 59 benchmark functions from CEC2014 and CEC2017. Experimental results demonstrate that MSDBO outperforms DBO, four advanced DBO variants, and several other popular algorithms in overall performance. Furthermore, MSDBO is employed for parameter identification in photovoltaic system models, further showcasing its effectiveness and reliability in real-world engineering applications.

Deep learning-based malaria parasite detection: convolutional neural networks model for accurate species identification of Plasmodium falciparum and Plasmodium vivax

Accurate malaria diagnosis with precise identification of Plasmodium species is crucial for an effective treatment. While microscopy is still the gold standard in malaria diagnosis, it relies heavily on trained personnel. Artificial intelligence (AI) advances, particularly convolutional neural networks (CNNs), have significantly improved diagnostic capabilities and accuracy by enabling the automated analysis of medical images. Previous models efficiently detected malaria parasites in red blood cells but had difficulty differentiating between species. We propose a CNN-based model for classifying cells infected by P. falciparum, P. vivax, and uninfected white blood cells from thick blood smears. Our best-performing model utilizes a seven-channel input and correctly predicted 12,876 out of 12,954 cases. We also generated a cross-validation confusion matrix that showed the results of five iterations, achieving 63,654 out of 64,126 true predictions. The model’s accuracy reached 99.51%, a precision of 99.26%, a recall of 99.26%, a specificity of 99.63%, an F1 score of 99.26%, and a loss of 2.3%. We are now developing a system based on real-world quality images to create a comprehensive detection tool for remote regions where trained microscopists are unavailable.

Identifying causes of aviation safety events using wW2V-tCNN with data augmentation

Identifying the causes of these safety events is crucial for safety agencies to create recommendations and for airlines to enhance procedures and mitigate hazards. This paper proposes a model to identify the causes of civil aviation safety events using a weighted Word2Vec-based Text-CNN (wW2V-tCNN) algorithm and data augmentation techniques. A corpus is built by matching narrative texts from investigation reports with cause labels from the Aviation Safety Network database. This corpus is transformed into Text-CNN inputs using a weighted sentence vector method based on word embeddings, considering word frequency and part-of-speech weighting. Additionally, a novel document balancing method is introduced for data augmentation. The proposed identification model achieves Macro-F1 and Macro-accuracy scores of 0.9803 and 0.9699, outperforming traditional methods and showing significant improvement over models like Doc2vec and SBERT. This model provides an accurate tool for safety agencies and airlines to analyze and effectively mitigate civil aviation safety events.