Journal of Computational Methods in Engineering Applications

Optimizing Transformer Models for Resource-Constrained Environments: A Study on Model Compression Techniques

Ziqian Luo — 2023-11-06

Recent progress in computer vision has been driven by transformer-based models, which consistently outperform traditional methods across various tasks. However, their high computational and memory demands limit their use in resource-constrained environments. This research addresses these challenges by investigating four key model compression techniques: quantization, low-rank approximation, knowledge distillation, and pruning. We thoroughly evaluate the effects of these techniques, both individually and in combination, on optimizing transformers for resource-limited settings. Our experimental findings show that these methods can successfully strike a balance between accuracy and efficiency, enhancing the feasibility of transformer models for edge computing.

Exploring the Factors of Supply Chain Concentration in Chinese A-Share Listed Enterprises

Yingda Tang — 2023-11-06

Using data of 1133 Chinese A-share listed enterprises on the Shanghai and Shenzhen Stock Exchanges for the period 2011–2022, this study examines the effects of corporate innovation, firm over-indebtedness, financing constraints, and firm profitability on supply chain concentration. The baseline results show that corporate innovation and profitability reduce supply chain concentration, while over-indebtedness and financing constraints increase it. These findings suggest that higher R&D investments and profitability enable firms to diversify their supply chains, whereas financial pressures lead to consolidation. The results remain robust after addressing endogeneity concerns using the system GMM approach. Heterogeneity analysis reveals stronger responses in large firms, state-owned enterprises, and high-tech industries. These results suggest policy implications of promoting R&D investments, reducing the debt levels, alleviating the financing constraints, and adopting profit-generating activities to diversify supply chain.

Deep Learning-Based Multifunctional End-to-End Model for Optical Character Classification and Denoising

Shuguang Xiong — 2023-11-15

Optical Character Recognition (OCR) has revolutionized document processing by converting scanned documents, PDFs, and images captured by cameras into editable and searchable text. This technology is crucial for digitizing historical documents, streamlining data entry processes, and improving accessibility for the visually impaired through text-to-speech technologies. Despite its widespread application, OCR faces significant challenges, especially in accurately recognizing text in noisy or degraded images. Traditionally, OCR systems have treated noise reduction and character classification as separate stages, which can compromise the overall effectiveness of text recognition. Our research introduces a groundbreaking Multifunctional End-to-End Model for Optical Character Classification and Denoising, which integrates these functions within a unified framework. By employing a dual-output autoencoder, our model concurrently denoises images and recognizes characters, thereby enhancing both the efficiency and accuracy of OCR. This paper outlines the model's development and implementation, explores the interplay between denoising and classification, and presents compelling experimental results that demonstrate marked improvements over conventional OCR methods.

CORAL-based Domain Adaptation Algorithm for Improving the Applicability of Machine Learning Models in Detecting Motor Bearing Failures

Guojun Zhang — 2023-11-03

Motor bearings are essential components in various industrial and transportation systems, vital for minimizing friction and enhancing machinery longevity. Failures in these bearings can lead to extensive machine downtime and significant repair costs, thereby emphasizing the need for effective predictive maintenance strategies. This paper focuses on leveraging advancements in Machine Learning (ML) and Artificial Intelligence (AI) to preemptively identify and rectify potential bearing failures, transitioning from traditional periodic maintenance to more efficient, condition-based approaches. We introduce a novel domain adaptation technique using Correlation Alignment (CORAL) to improve the accuracy of fault predictions across different operational settings. This method effectively minimizes the statistical disparities between training and operational data, enhancing the adaptability and effectiveness of predictive models. The results indicate that models equipped with domain adaptation outperform traditional models, particularly in their ability to generalize across diverse environments, thereby supporting more reliable and efficient predictive maintenance practices. This research contributes to the ongoing evolution of maintenance strategies in industrial settings, highlighting the potential of AI to transform traditional practices by reducing unexpected downtime and optimizing maintenance schedules.

AI-Driven Health Advice: Evaluating the Potential of Large Language Models as Health Assistants

Yanlin Liu — 2023-11-06

This study aims to evaluate whether the GPT model can be a health assistant by addressing health concerns from three aspects: providing preliminary guidance, clarifying information, and offering accessible recommendations. 31 questions in total were collected from multiple online health platforms, which included diverse health concerns across different age ranges and genders. A tailored system prompt was built to guide GPT model GPT-3.5-turbo generating responses. The evaluation metrics are designed based on 3 metrics: “Preliminary Guidance”, “Clarifying Information”, and “Accessibility and Convenience”, which is used to evaluate responses with score method from 0 to 5. Lastly, the generated responses were evaluated using established metrics by an experienced medical doctor with over 20 years of experience in the fields of general and preventive care. The results indicate that LLMs demonstrated moderate performance in both the ‘preliminary guidance’ and ‘clarifying information’ aspects. Specifically, the mean score for ‘preliminary guidance’ was 3.65, implying that LLMs are capable of offering valuable insights when symptoms indicate the need for urgent or emergency care, as well as providing reassurance to patients for minor symptoms. In a similar manner, the mean score for ‘clarifying information’ was 3.87, demonstrating that LLMs effectively provide supplementary information to aid patients in making informed decisions. However, the mean score for ‘accessibility and convenience’ was notably lower at 2.65, highlighting a deficiency in LLMs’ ability to offer advice customized to the specific needs of individual patients.

Efficient Bank Fraud Detection with Machine Learning

Rong Zhang — 2023-10-23

The rapid expansion of IT technology has led to a significant increase in financial transactions, accompanied by a corresponding rise in fraudulent activities. This paper tackles the challenge of detecting fraud in various forms, such as credit card fraud, banker cheque fraud, and online funds transfer fraud, which have become increasingly sophisticated. Traditional methods struggle to keep pace with these evolving fraud techniques, necessitating advanced approaches. We propose the use of machine learning algorithms to enhance the detection of fraudulent transactions. Utilizing the BankSim dataset from Kaggle, which includes features like age, gender, payment domain, and transaction amount, we conducted a comprehensive analysis. The dataset was preprocessed to handle missing values and balance the instances of fraud. We then applied several machine learning algorithms, including K-Nearest Neighbors (KNN), Naive Bayes, and Support Vector Machine (SVM), training these models on a training set and evaluating them on a test set. The performance of these models was assessed using precision, recall, and F1-measure metrics. Our findings demonstrate that the SVM algorithm achieved the highest accuracy at 99.23%, significantly outperforming the other algorithms and previous studies. This study highlights the effectiveness of machine learning, particularly SVM, in developing robust fraud detection systems, offering a promising solution to improve financial security.

Enhanced E-Commerce Sales Forecasting Using EEMD-Integrated LSTM Deep Learning Model

Yunxiang Gan — 2023-11-11

E-commerce sales data often exhibit complex time series characteristics and are influenced by multiple factors, making traditional forecasting methods inadequate in capturing these dynamics. To address these challenges, this paper presents a forecasting model that integrates Ensemble Empirical Mode Decomposition (EEMD) with Long Short-Term Memory (LSTM) networks. The model first applies EEMD to decompose the original data signal into multiple Intrinsic Mode Function (IMF) components. These components, along with the original data, are then fed into the LSTM network for predictive analysis. As a case study, the proposed model is tested using a sales dataset of an Amazon clothing product. The results demonstrate that the model achieves a forecasting accuracy of 91%, surpassing several commonly used forecasting approaches in precision and reliability. This study highlights the potential of the EEMD-LSTM approach in improving sales forecasts for e-commerce platforms.

Machine Learning-Based System Reliability Analysis with Gaussian Process Regression

Lisang Zhou — 2023-11-13

Machine learning-based reliability analysis methods have shown great advancements for their computational efficiency and accuracy. Recently, many efficient learning strategies have been proposed to enhance the computational performance. However, few of them explores the theoretical optimal learning strategy. In this article, we propose several theorems that facilitates such exploration. Specifically, cases that considering and neglecting the correlations among the candidate design samples are well elaborated. Moreover, we prove that the well-known U learning function can be reformulated to the optimal learning function for the case neglecting the Kriging correlation. In addition, the theoretical optimal learning strategy for sequential multiple training samples enrichment is also mathematically explored through the Bayesian estimate with the corresponding lost functions. Simulation results show that the optimal learning strategy considering the Kriging correlation works better than that neglecting the Kriging correlation and other state-of-the art learning functions from the literatures in terms of the reduction of number of evaluations of performance function. However, the implementation needs to investigate very large computational resource.

Exploring the Path of Transformation and Development for Study Abroad Consultancy Firms in China

Ping Ren — 2023-11-20

In recent years, with the changing landscape of international education and the growing demand from Chinese students, study abroad consultancy firms in China need to adopt transformational development strategies to address challenges and maintain competitiveness. This study investigated the relationships between key performance indicators and several factors through a questionnaire survey of 158 consultancy firms. The factors examined included service diversification, technology adoption, talent management, and regulatory compliance. Descriptive statistical analysis was employed to analyze the data. The results showed that service scope diversification was positively correlated with firm performance. Technology adoption was positively correlated with operational efficiency. Talent management was positively correlated with service quality. Regulatory compliance was positively correlated with firm reputation. Consultancy firms that took progressive approaches in diversifying services, adopting new technologies, cultivating talent, and ensuring compliance demonstrated superior performance, efficiency, quality, and reputation compared to their less innovative counterparts. This research provides empirical evidence to support the transformation of Chinese study abroad consultancy firms. It also highlights the need for future studies to consider causality and contextual variations to gain deeper insights into this issue.