Neural Network Model Research Articles

The Mechanism of Employee Characteristics on Promotion: Building the Predictive Machine Learning Model

Purpose: This study developed a predictive model based on employee characteristics, analyzes multidimensional data to identify predictive factors of promotion outcomes, and provides data-driven insights for organizations to improve talent management. Methods: Decision tree, random forest, and neural network models were built on the existing dataset (N=54,808). Then, the models predictive efficacies were evaluated to select the optimal model using several methods. Results: The accuracy of decision tree was 85.7%. The accuracy of artificial neural network and random forest model was 93%. For promoted class, random forest had higher precision, while neural network performed slightly better in recall and F1 score. For non-promoted class, both models perform almost identically in precision, recall, and F1 score. Conclusion: Random forest and neural network had good predictive efficacy for employee promotion. If avoiding false promotions is more important (i.e., precision matters more), random forest is the better model. If capturing more promotions is critical (i.e., recall matters more), even at the cost of some false positives, then the neural network is slightly better.

Research on Automatic Detection of Citrus Pests and Diseases Using Deep Neural Network Model

Citrus fruit diseases are a major cause of citrus yield decline. Deep learning methods have achieved promising results in many artificial intelligence problems, which motivates us to apply them to the challenge of recognizing citrus pests and diseases. In this paper, two progressive convolutional neural network (CNN) models are proposed. The proposed CNN models are designed to distinguish between healthy fruits and common citrus pest fruits. The first model is a four-branch parallel CNN based on a feature fusion (ODC-FFN) structure for the automatic detection of citrus diseases. The model takes the size, color, shape, and texture images of citrus surface diseases as inputs and uses a fusion feature extractor to design the four-branch network structure. Through comparison experiments with other models, its recognition accuracy reaches 94.68%. The second model is based on the first model, the multiple attention-based citrus disease classification model (ODC-MAN) is constructed by embedding spatial attention and channel attention mechanisms as feature extractors, and the convolutional layer fusion is changed to classification layer fusion, and the integrated network model is further constructed through integrated learning. The recognition accuracy is further improved to 98.21% by comparison experiments, which indicates that the integrated model using classification layer fusion is effective.

Estimation of Evaporation Using Long Short Term Memory and Gated Recurrent Unit Based Neural Network

This study presents the comparison of conventional Artificial Neural Network (ANN) and advanced neural networks to predict weekly potential evaporation for Anand, Gujarat, India, which comes under the subtropical climatic zone. Recently, many advanced deep neural structures like Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) have been introduced that have excellent prediction accuracy. Climate data such as bright sunshine hours, rainfall, wind speed, maximum and minimum temperature, and maximum and minimum relative humidity have been used to train and test the conventional and advanced neural network models. A comparison was made for the estimation of evaporation predicted by these models. The performance results show that deep neural network models with advanced structures like LSTM and GRU have performed better in terms of Root Mean Square Error and correlation coefficient and are able to learn the events very well in comparison to conventional neural network structures.

Abstract 6566: FTIR microscope imaging and neural network analysis reveal consistent spectral markers of cancer-associated fibroblasts in cells, plasma sEVs, and tissues

Background: Cancer-associated fibroblasts (CAFs) are pivotal contributors to tumor progression, influencing tumor biology through interactions within the tumor microenvironment. Distinguishing CAFs from normal fibroblasts is essential for understanding their molecular mechanisms and identifying relevant biomarkers. This study utilized Fourier-transform infrared (FTIR) microscope imaging spectra and neural networks (NNs) to classify CAFs and normal fibroblasts and explored the consistency of identified spectral markers across blood plasma small extracellular vesicles (sEVs) and tissue samples. Methods: FTIR spectra from CAFs and normal fibroblasts were preprocessed using normalization and baseline correction. A feed-forward NN model was developed, employing regularization techniques to reduce overfitting. Model performance was evaluated using 10-fold cross-validation, with metrics including accuracy, area under the curve (AUC), sensitivity, and specificity. NN-based sensitivity analysis identified key infrared (IR) spectral regions for classification. Comparative FTIR analysis of blood plasma sEVs and tissue samples was conducted to assess the reproducibility of identified spectral markers. Results: The NN model demonstrated high classification performance, achieving 92% accuracy, an AUC of 0.89, and robust sensitivity and specificity metrics. Sensitivity analysis highlighted significant IR spectral regions, including Amide I and Amide II (proteins): Indicative of changes in protein structure and extracellular matrix remodeling. Lipid Regions: Reflecting altered lipid metabolism, critical for membrane dynamics and signaling in CAFs. Nucleic Acid-Associated Regions: Suggesting transcriptional and epigenetic modifications linked to fibroblast activation. Notably, similar spectral markers were observed in FTIR analysis of blood plasma sEVs and tissue samples, revealing a consistent pattern associated with CAFs. This suggests systemic changes in the tumor microenvironment that are detectable across different biological matrices. Conclusion: This study highlights the potential of NN-based FTIR spectral analysis as a robust and versatile tool for identifying biomarkers of CAFs. The reproducibility of key spectral markers in blood plasma sEVs and tissue samples underscores the translational potential of this approach for cancer diagnostics and therapeutic targeting of the tumor microenvironment. Citation Format: Abhay Mishra, Akhilesh Mishra, Saroj Kumar. FTIR microscope imaging and neural network analysis reveal consistent spectral markers of cancer-associated fibroblasts in cells, plasma sEVs, and tissues [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6566.

Abstract 2708: Prediction of radiosensitivity in glioblastoma tumors after standard and low dose rates of ionizing radiation

Glioblastomas (GBM) exhibit high infiltration into the brain parenchyma, impairing surgical resection and leading to recurrence. After standard of care, including tumor resection, radiation therapy (RT) and chemotherapy, most tumors recur at the tumor margins. These tumors show an increased aggressiveness and therapeutic resistance. The underlying mechanisms for these cellular transitions are unclear. Research suggests that RT paradoxically alters the GBM microenvironment promoting cancer cell survival upon recurrence. Some of these alterations in the irradiated tumor microenvironment include increased oxidative stress, hypoxia, cellular senescence, vasculogenesis and ECM compositional and structural changes.We use an ex vivo platform that combines the hallmarks of GBM tumors: high cellularity, the neural circuitry that surrounds tumors, a hydrogel-based ECM that mimics the biophysical and biochemical properties of GBM and the intermittent hypoxic environment present in native tumors, to characterize and evaluate the therapeutic benefits of RT and the underlying mechanisms of neurotoxicity to more effectively treat the tumor while minimizing the deleterious side effects.Our results show that radiation treatment increases glycolysis in tumor cells and that immediate response to radiation produces a more significant ATP production decrease in hypoxia environments. We have also investigated the long-term energy production and mitochondrial dysfunction in the tumor cells at different radiation rates and under low oxygen environments. Interestingly, the maximal respiration capacity in GBM cells increases with the presence of conditioned media from neuronal cultures in normoxic conditions but has the opposite effect in hypoxic environments.Proliferation ability of cells on stiffer matrices is greater than that on softer matrices. In the presence of an altered biomechanical microenvironment, a stiffer tumor tissue is more prone to immunosuppression and epithelial-mesenchymal transition. After cells are irradiated, the cell migration is different with different matrix stiffness. Moreover, softer matrices are more likely to promote cellular radiosensitivity, while stiffer matrices can inhibit cellular radiosensitivity.The use of preclinical tools that can model key features of the GBM and normal brain tissue can give a better understanding of the spatial-temporal characteristics of tumor microenvironment alterations during progression and response to therapy. Citation Format: Isabella Rivera, Ryan Yao, Kim Selting, Catherine Best-Popescu, Brendan Harley, Sara Pedron-Haba. Prediction of radiosensitivity in glioblastoma tumors after standard and low dose rates of ionizing radiation [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2708.

Development and Sharing of Traditional Cultural and Educational Resources Based on the Internet of Things

Integrating traditional, cultural and educational resources into modern learning platforms has become a resilient need in an increasingly globalized and digitally connected world. While technology continues to evolve at a rapid pace, the preservation and dissemination of cultural heritage remain crucial for maintaining the unique identities of societies. Research explores the integration of internet of things (IoT) technologies and advanced deep learning (DL) methods to manage and distribute data that include digital representations of artifacts, historical sites, and cultural practices. Data including location, environmental conditions, and user interactions, are collected via IoT devices to enhance the learning experience. Pre-processing techniques, such as median filtering and standardization, are applied to clean and normalize the data, ensuring its quality and consistency. Feature extraction methods, including discrete wavelet transform (DWT) and principal component analysis (PCA), are employed to capture the most relevant characteristics of the cultural data. The novel refined nutcracker optimized deep neural network (RNO-DNN) model is designed to intelligently analyze the pre-processed data, enabling personalized learning and immersive cultural experiences. The RNO-DNN enhances the accuracy of content recommendations and optimizes the system’s performance in real-time interactions. It also addresses challenges such as data privacy, computational efficiency, and the technical infrastructure required for RNO-DNN integration. The findings highlight the potential of combining IoT with DL to revolutionize the accessibility, engagement, and preservation of traditional cultural content, making it a critical tool for modern education and cultural heritage management.

A Comparative Study of VAR and ANN Models for the Forecasting of Eggplant Wholesale Prices in Lucknow, India

Accurate forecasting of agricultural commodity prices, particularly vegetables, is essential for comprehending market dynamics and maintaining economic stability among stakeholders. Effective price forecasting fosters agricultural sustainability and economic resilience. This study aims to identify the optimal model for forecasting wholesale eggplant prices in the Lucknow market, comparing the traditional vector autoregressive (VAR) model with the Artificial Neural Network (ANN) model. The most commonly used evaluation metrics were applied to compare and assess model performance. For model building, 204 monthly observations from January 2008 to December 2024, comprising a dataset of wholesale prices and total arrival of eggplant in the Lucknow center, were utilized. In the modeling process, the entire dataset was partitioned into two subsets, training and testing data, with a split of 80:20. To check stationarity, the Augmented Dickey-Fuller test was conducted, and the lag order for the VAR model was selected based on the minimum AIC value, with [5] as the optimal lag order. Similarly, the ANN model was created using [5] lags of the variables, as independent variables. The findings indicate that the ANN [10: 3: 2: 1] model surpasses the VAR (5) model in forecasting wholesale prices within the test dataset. However, both models exhibit overfitting, likely attributable to the VAR model's inadequacy in capturing the heteroscedasticity effect and the ANN model's dependence on a limited dataset and the number of lag variables incorporated as inputs. These methods can be applied to model and forecast prices of other agricultural commodities and have potential in broader agricultural research, as their utilization in this field remains limited.

Stock Price and Nifty 50 Index Forecasting Using LNN and BiLSTM-Based Deep Learning Models

This paper presents two complementary deep learning approaches for financial market forecasting in the Indian context: a Liquid Neural Network (LNN) model for individual stock price prediction and a Bidirectional Long Short-Term Memory (BiLSTM) model for NIFTY 50 index forecasting. The LNN model leverages time-varying parameters to predict stock prices for a 7-day horizon, adapting dynamically to market volatility and temporal patterns. For NIFTY 50 index prediction, we propose a novel hybrid approach combining technical indicators (OHLC-derived) with financial fundamentals of constituent companies. Principal Component Analysis (PCA) is applied to extract significant financial features from each NIFTY 50 company, which are then merged using a union-based approach and combined with technical indicators to train the BiLSTM model.Both implementations demonstrate robustness across various market conditions, including high-volatility periods and significant market events. Stock Price Prediction, Nifty 50 Forecasting, Liquid Neural Network (LNN), Bidirectional LSTM (BiLSTM), Deep Learn- ing, Technical Indicators, Financial Fundamentals, PCA, Time Series Forecasting, Indian Stock Market.

Interpretable AI-assisted clinical decision making for treatment selection for brain metastases in radiation therapy.

AI modeling CDM can improve the quality and efficiency of clinical practice or provide secondary opinion consultations for patients with limited medical resources to address healthcare disparities. In this study, we developed an interpretable AI model to select radiotherapy treatment options, that is, whole-brain radiation therapy (WBRT) versus stereotactic radiosurgery (SRS), for patients with brain metastases. A total of 232 patients with brain metastases treated by radiation therapy from 2018 to 2023 were obtained. CT/MR images with contoured target lesions and organs-at-risk (OARs) as well as non-image-based clinical parameters were extracted and digitized as inputs to the model. These parameters included (1) tumor size, shape, location, and proximity of lesions to OARs; (2) age; (3) the number of brain metastases; (4) Eastern Cooperative Oncology Group (ECOG) performance status; (5) presence of neurologic symptoms; (6) if surgery was performed (either pre/post-op RT); (7) newly diagnosed cancer with brain metastases (de-novo) versus re-treatment (either local or distant in the brain); (8) primary cancer histology; (9) presence of extracranial metastases; (10) extent of extracranial disease (progression vs. stable); and (11) receipt of systemic therapy. One vanilla and two interpretable 3D convolutional neural networks (CNN) models were developed. The vanilla one-path model (VM-1) uses only images as input, while the two interpretable models use both images and clinical parameters as inputs with two (IM-2) and 11 (IM-11) independent paths, respectively. This novel design allowed the model to calculate a class activation score for each input to interpret its relative weighting and importance in decision-making. The actual radiotherapy treatment (WBRT or SRS) used for the patients was used as ground truth for model training. The model performance was assessed by Stratified-10-fold cross-validation, with each fold consisting of selected 184 training, 24 validation, and 24 testing subjects. A total of 232 brain metastases patients treated by WBRT or SRS were evaluated, including 80 WBRT and 152 SRS patients. Based on the images alone, the VM-1 model prescribed correctly for 143 (94%) SRS and 67 (84%) WBRT cases. Based on both images and clinical parameters, the IM-2 model prescribed correctly for 149 (98%) SRS and 74 (93%) WBRT cases. IM-11 provided the most interpretability with a relative weighting for each input as follows: CT image (59.5%), ECOG performance status (7.5%), re-treatment (5%), extracranial metastases (1.5%), number of brain metastases (9.5%), neurologic symptoms (3%), pre/post-surgery (2%), primary cancer histology (2%), age (1%), progressive extracranial disease (6%), and receipt of systemic therapy (4.5%), reflecting the importance of all these inputs in clinical decision-making. Interpretable CNN models were successfully developed to use CT/MR images and non-image-based clinical parameters to predict the treatment selection between WBRT and SRS for brain metastases patients. The interpretability makes the model more transparent, carrying profound importance for the prospective integration of these models into routine clinical practice, particularly for informing real-time clinical decision-making.

AI Attacks AI: Recovering Neural Network Architecture from NVDLA Using AI-Assisted Side Channel Attack

During the last decade, there has been a stunning progress in the domain of Artificial Intelligence (AI) aided by highly trained Machine Learning (ML) models. Such models are valuable Intellectual Property (IP) and, therefore, have been subjected to various model recovery attacks. In this work, we study the vulnerabilities of commercial, open-source accelerator NVDLA and present the first successful model recovery attack. For this purpose, we used power and timing information from the side-channel leakage of convolutional neural networks (CNN) models to train CNN-based attack models. Utilizing these attack models, we demonstrate that even with a highly pipelined architecture, multiple parallel execution in the accelerator along with Linux OS running tasks in the background, recovery of number of layers, kernel sizes, output neurons and distinguishing different layers, is possible with very high accuracy. This is also the first work to show the impact of differences in hyperparameters on the power traces. Our solution is fully automated, AI-based, and portable to other hardware neural networks, thus presenting a greater threat toward IP protection. Using LeNet as the target victim model, we demonstrate an accuracy of more than 95% in recovering various parameters. This study presents a serious practical threat, in the form of side-channel attack, towards complex commercial architectures. Furthermore, we show that AI-guided attack significantly boosts the attacker capability.

Rock Joint Detection from Borehole Imaging Logs Using a Convolutional Neural Networks Model

Rock Joint Detection from Borehole Imaging Logs Using a Convolutional Neural Networks Model

Design of BCC/FCC dual-solid solution refractory high-entropy alloys through CALPHAD, machine learning and experimental methods

Refractory high-entropy alloys (RHEAs) typically exhibit a body-centered cubic (BCC) structure with excellent strength but poor ductility, which limits their practical applications. In this study, we designed BCC/FCC dual-phase RHEAs through phase diagram calculations and neural network modeling. The analysis of the binary phase formation relationships among alloying elements enabled the preliminary screening and inclusion of 13 liquid-phase-separated BCC/FCC dual-phase RHEAs in the training dataset for the machine learning model. Two strategic binary classifications of this dataset were conducted on HEAs to identify their “multiphase” and “solid solution” structures. Consequently, two neural network models were trained, achieving accuracies of 89.52% and 89.83%, respectively. These models predicted 51 BCC/FCC dual-phase RHEAs among 504 novel RHEAs, representing the first successful compositional design of metastable BCC/FCC dual-phase RHEAs. The arc-melted alloys exhibited refined dendritic structure. This study provides valuable insights for the tailored design of novel multi-phase RHEAs to achieve specific targeted properties.

Artificial Neural Network Model for Evaluating Load Capacity of RC Deep Beams

Using artificial neural networks (ANN), numerous models were developed for predicting the ultimate shear strength of reinforced concrete deep beams. Many experimental result databases from earlier research were carefully gathered for this study. Two hundred fifty-three findings from experiments were included in this database. The ultimate shear strength was the output parameter, while ten factors were determined as input parameters for the ANN model based on the completed literature research. The required model was constructed using a back propagation neural network. The model of the neural networks was determined using the trial-and-error method. It was discovered that, inside the range of the input boundaries considered, the ANN model could accurately estimate the ultimate shear strength of deep beams. The measured shear strength and the shear strength predicted by the ANN model have a high correlation coefficient of 0.97, indicating a strong relationship between the predicted and actual values. The results show that, given the range of input parameters, ANN offers an excellent agreement of interest as a practical technique for estimating the ultimate shear strength. A parametric investigation was performed using the trained neural network model to assess how the input parameters affected the shear strength capacity of deep beams.

Efficient Machine Learning for Malware Detection

As the landscape of cyber threats continues to expand, malware detection has become increasingly crucial for maintaining robust cybersecurity. While standard malware detection techniques such as signature-based methods are very effective and widespread, they face certain challenges with zero-day and novel malware. The emergence of artificial intelligence in recent years has led to the development of alternative approaches to this issue, specifically through machine learning techniques. This research aims to analyze the effectiveness and viability of one such machine learning approach; the use of a Convolutional Neural Network (CNN) model for the classification of benign and malicious Windows executable binaries. To accomplish this, we gathered a substantial dataset of both benign and malicious Windows binaries and converted them into grayscale images to train several CNN models with slightly varying architecture for the classification task. Following the training of the models, they were evaluated on an unseen test dataset to compare label predictions against each other, as well as Windows Defender. This approach aims to achieve a definitive metric for determining the effectiveness of this type of malware detection for Windows-based antivirus applications. What we found is that certain CNN models are not only able to perform on par with Windows Defender, but in some cases even outperform them. In conclusion, our study demonstrated that utilizing CNN models with grayscale image conversion of Windows binaries is an effective and efficient approach to malware detection.

Smart Application of Deep Learning Technology in Bacterial Security Identification

The rapid development of artificial intelligence technology has promoted the intelligence and automation of smart security systems, making security monitoring more efficient and accurate. Based on computer vision and natural language processing technology, smart security covers key technical fields such as face recognition, behavior monitoring and intelligent inspection. Although artificial intelligence technology has brought revolutionary changes to smart security, it still faces challenges such as privacy security, misidentification and missed identification, which need to be continuously improved and resolved. This paper deeply explores the application and development of artificial intelligence technology in smart security, and provides an important reference for future research and practice in the field of smart security. Bacteria are widely present in nature, and accurate identification of bacterial types is crucial in clinical diagnosis, food safety monitoring and other fields. However, traditional bacterial identification methods based on microscopy observation have many defects, are time-consuming and labor-intensive, and rely on manual experience. This paper explores methods for intelligent bacterial identification based on computer vision and deep learning technology. First, the application principles of deep learning in the field of image processing are introduced, especially the advantages of convolutional neural networks in bacterial image feature extraction. Then, the latest research progress of bacterial identification technology based on deep learning is reviewed, and the performance of various convolutional neural network models in bacterial classification tasks is commented on. Finally, the main challenges currently faced in this field are pointed out, providing a theoretical basis for promoting bacterial identification technology based on deep learning.

A Multi-Scale Fusion Convolutional Network for Time-Series Silicon Prediction in Blast Furnaces

In steel production, the blast furnace is a critical element. In this process, precisely controlling the temperature of the molten iron is indispensable for attaining efficient operations and high-grade products. This temperature is often indirectly reflected by the silicon content in the hot metal. However, due to the dynamic nature and inherent delays of the ironmaking process, real-time prediction of silicon content remains a significant challenge, and traditional methods often suffer from insufficient prediction accuracy. This study presents a novel Multi-Scale Fusion Convolutional Neural Network (MSF-CNN) to accurately predict the silicon content during the blast furnace smelting process, addressing the limitations of existing data-driven approaches. The proposed MSF-CNN model extracts temporal features at two distinct scales. The first scale utilizes a Convolutional Block Attention Module, which captures local temporal dependencies by focusing on the most relevant features across adjacent time steps. The second scale employs a Multi-Head Self-Attention mechanism to model long-term temporal dependencies, overcoming the inherent delay issues in the blast furnace process. By combining these two scales, the model effectively captures both short-term and long-term temporal dependencies, thereby enhancing prediction accuracy and real-time applicability. Validation using real blast furnace data demonstrates that MSF-CNN outperforms recurrent neural network models such as Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU). Compared with LSTM and the GRU, MSF-CNN reduces the Root Mean Square Error (RMSE) by approximately 22% and 21%, respectively, and improves the Hit Rate (HR) by over 3.5% and 4%, highlighting its superiority in capturing complex temporal dependencies. These results indicate that the MSF-CNN adapts better to the blast furnace’s dynamic variations and inherent delays, achieving significant improvements in prediction precision and robustness compared to state-of-the-art recurrent models.

Enhancing water saturation predictions from conventional well logs in a carbonate gas reservoir with a hybrid CNN-LSTM model

Predicting water saturation in the South Pars gas field’s carbonate reservoirs has been challenging due to limitations in conventional methods like the Archie equation. This leads to unproduced gas volume and water coning phenomena, leaving a gap for an intelligent model to determine the water saturation profile more accurately. Four machine learning algorithms: XGBoost, long short-term memory (LSTM), 1-dimensional convolutional neural network (1D-CNN), and a hybrid CNN-LSTM were developed to predict experimental water saturation values from conventional well logs. A total of 10,674 data points were collected from four wells in the South Pars gas field, where well logging evaluations and core measurements were available. The methodology includes data pre-processing to de-noise and transform data into a proper format, feature engineering followed by feature selection based on domain knowledge, models’ training, testing, and optimizing. XGBoost hyperparameters were tuned using grid search, and each neural network model was optimized using a genetic algorithm. The network architectures were designed with specific layers and activation functions to enhance predictive capabilities. A robust validation process was conducted using a blind well dataset. Archie equation was also applied for better comparative analysis. Results showed that the CNN-LSTM model was the most effective, with an R-squared accuracy score of 0.859, while the Archie equation had the least R-squared accuracy of 0.322 in the validation dataset. Furthermore, utilizing an inclusive dataset, a new CNN-LSTM model was constructed with a 10-fold framework, achieving an average R-squared of 0.968. This model was utilized to develop an application that displays the reservoir’s water saturation profile and other features as a well log report within a dashboard. The findings support two points: first, 1D-CNN and LSTM work better together in a hybrid structure, and second, the Archie equation is ineffective in carbonate reservoirs of the South Pars gas field. The study intends to enrich the literature by employing a hybrid machine learning algorithm compared to earlier studies and surpassing traditional protocols by comparing algorithms in a blind well to offer a reliable validation mechanism.Graphical abstract

Comparative Analysis of Machine Learning and Deep Learning Models for Lung Cancer Prediction Based on Symptomatic and Lifestyle Features

Lung cancer remains a leading cause of global mortality, with early detection being critical for improving the patient survival rates. However, applying machine learning and deep learning effectively for lung cancer prediction using symptomatic and lifestyle data requires the careful consideration of feature selection and model optimization, which is not consistently addressed in the existing research. This research addresses this gap by systematically evaluating and comparing the predictive efficacy of several machine learning and deep learning models, employing rigorous data preprocessing, including feature selection with Pearson’s correlation, outlier removal, and normalization, on a patient symptom and lifestyle factor dataset from Kaggle. Machine learning classifiers, including Decision Trees, K-Nearest Neighbors, Random Forest, Naïve Bayes, AdaBoost, Logistic Regression, and Support Vector Machines, were implemented using Weka simultaneously with neural network models with 1, 2, and 3 hidden layers, which were developed in Python within a Jupyter Notebook environment. The model performance was assessed using K-fold cross-validation and 80/20 train/test splitting. The results highlight the importance of feature selection for enhancing the model accuracy and demonstrate that the single-hidden-layer neural network, trained for 800 epochs, achieved a prediction accuracy of 92.86%, outperforming the machine learning models. This study contributes to developing more effective computational methods for early lung cancer detection, ultimately supporting improved patient outcomes.

An analytic hierarchy process combined with artificial neural network model to evaluate sustainable sludge treatment scenarios.

An analytic hierarchy process combined with artificial neural network model to evaluate sustainable sludge treatment scenarios.

Research on Wearable Devices for Pedestrian Navigation Based on the Informer Model Zero-Velocity Update Architecture

When natural disasters such as earthquakes occur, accurate navigation and positioning information may not be available, making a purely inertial pedestrian navigation system particularly important for rescuers. In this paper, researchers propose a zero-velocity update architecture for pedestrian navigation based on the Informer model, which is integrated into wearable devices. This architecture modifies the fully connected layer of the Informer model to be used for the binary classification task of the zero-velocity update method (ZUPT), allowing for accurate identification of gait information at each moment using only inertial measurement data. By wearing the device on the foot during natural disasters like earthquakes, the location of the pedestrian can be more accurately determined, facilitating rescue efforts. During the experimental process, a Kalman filter model was constructed to achieve zero-velocity updating of the pedestrian’s motion trajectory. A 2000 m walking experiment and a 210 m mixed-gait experiment were conducted to accurately identify gait information at each moment, thereby reducing the cumulative error of the inertial system. Subsequently, a convolutional neural network (CNN) model and a model combining CNN with a long short-term memory network (CNN + LSTM) were introduced as comparative experiments to verify the performance of the proposed architecture. The experimental results demonstrate that the proposed architecture enhances the adaptability of the zero-velocity update algorithm in underground or sheltered spaces, with all results outperforming the other two models.