Search results

This study aims to construct a sentiment series generation method for danmu comments based on deep learning, and explore the features of sentiment series after clustering.

This study consisted of two main parts: danmu comment sentiment series generation and clustering. In the first part, the authors proposed a sentiment classification model based on BERT fine-tuning to quantify danmu comment sentiment polarity. To smooth the sentiment series, they used methods, such as comprehensive weights. In the second part, the shaped-based distance (SBD)-K-shape method was used to cluster the actual collected data.

The filtered sentiment series or curves of the microfilms on the Bilibili website could be divided into four major categories. There is an apparently stable time interval for the first three types of sentiment curves, while the fourth type of sentiment curve shows a clear trend of fluctuation in general. In addition, it was found that “disputed points” or “highlights” are likely to appear at the beginning and the climax of films, resulting in significant changes in the sentiment curves. The clustering results show a significant difference in user participation, with the second type prevailing over others.

Their sentiment classification model based on BERT fine-tuning outperformed the traditional sentiment lexicon method, which provides a reference for using deep learning as well as transfer learning for danmu comment sentiment analysis. The BERT fine-tuning–SBD-K-shape algorithm can weaken the effect of non-regular noise and temporal phase shift of danmu text.

This paper proposes a solution for recognizing human faces under mask-wearing. The lower part of human face is occluded and could not be used in the learning process of face recognition. So, the proposed solution is developed to recognize human faces on any available facial components which could be varied depending on wearing or not wearing a mask.

The proposed solution is developed based on the FaceNet framework, aiming to modify the existing facial recognition model to improve the performance of both scenarios of mask-wearing and without mask-wearing. Then, simulated masked-face images are computed on top of the original face images, to be used in the learning process of face recognition. In addition, feature heatmaps are also drawn out to visualize majority of parts of facial images that are significant in recognizing faces under mask-wearing.

The proposed method is validated using several scenarios of experiments. The result shows an outstanding accuracy of 99.2% on a scenario of mask-wearing faces. The feature heatmaps also show that non-occluded components including eyes and nose become more significant for recognizing human faces, when compared with the lower part of human faces which could be occluded under masks.

The convolutional neural network based solution is tuned up for recognizing human faces under a scenario of mask-wearing. The simulated masks on original face images are augmented for training the face recognition model. The heatmaps are then computed to prove that features generated from the top half of face images are correctly chosen for the face recognition.

The growing availability of naturalistic driving datasets (NDDs) presents a valuable opportunity to develop various models for autonomous driving. However, while current NDDs include data on vehicles with and without intended driving behavior changes, they do not explicitly demonstrate a type of data on vehicles that intend to change their driving behavior but do not execute the behaviors because of safety, efficiency, or other factors. This missing data is essential for autonomous driving decisions. This study aims to extract the driving data with implicit intentions to support the development of decision-making models.

According to Bayesian inference, drivers who have the same intended changes likely share similar influencing factors and states. Building on this principle, this study proposes an approach to extract data on vehicles that intended to execute specific behaviors but failed to do so. This is achieved by computing driving similarities between the candidate vehicles and benchmark vehicles with incorporation of the standard similarity metrics, which takes into account information on the surrounding vehicles' location topology and individual vehicle motion states. By doing so, the method enables a more comprehensive analysis of driving behavior and intention.

The proposed method is verified on the Next Generation SIMulation dataset (NGSim), which confirms its ability to reveal similarities between vehicles executing similar behaviors during the decision-making process in nature. The approach is also validated using simulated data, achieving an accuracy of 96.3 per cent in recognizing vehicles with specific driving behavior intentions that are not executed.

This study provides an innovative approach to extract driving data with implicit intentions and offers strong support to develop data-driven decision-making models for autonomous driving. With the support of this approach, the development of autonomous vehicles can capture more real driving experience from human drivers moving towards a safer and more efficient future.

This study aims to examine the influence of training and development (TAD), including performance appraisal (PFA), on the performance of microfinance institutions in a developing economy.

A random sampling drew 100 microfinance institutions in Ghana's Greater Accra Region. Then, a purposive sampling approach selected a cross-section of employees in these institutions. Finally, the data were collected from a sample of 367 respondents, such as managers, utilizing a survey questionnaire. Structural equation modeling (SEM) was used to test hypothesized relationships.

The study results indicate that PFA has a statistically significant positive relationship with organizational performance, and this relationship is partially mediated by job satisfaction (JBS). Interestingly, the TAD process does not have a statistically significant positive relationship with organizational performance when JBS is present in the model. In fact, JBS fully mediates the relationship between TAD and organizational performance.

The study is limited to microfinance in tiers two and three in Greater Accra city of Ghana and did not include the entire country. Although the city of Accra provides a generalized representation of the research, which can be replicated, some variables and results may be impacted if other tiers of microfinance organizations are incorporated.

TAD, as well as PFA, enhance the performance of microfinance and can be utilized as tools for competitive advantage in small and medium-sized enterprises (SMEs) (e.g. microfinance institutions). The study accentuates the value of TAD, PFA and JBS in microfinance in a developing country like Ghana.

This is an original study investigating the effect of TAD and PFA practices on the performance of SMEs in a developing country like Ghana. Also, the study analyses JBS as a mediation variable to performance using SEM, which advances the research methodology in this research field.

COVID-19 continues to spread, and cause increasing deaths. Physicians diagnose COVID-19 using not only real-time polymerase chain reaction but also the computed tomography (CT) and chest x-ray (CXR) modalities, depending on the stage of infection. However, with so many patients and so few doctors, it has become difficult to keep abreast of the disease. Deep learning models have been developed in order to assist in this respect, and vision transformers are currently state-of-the-art methods, but most techniques currently focus only on one modality (CXR).

This work aims to leverage the benefits of both CT and CXR to improve COVID-19 diagnosis. This paper studies the differences between using convolutional MobileNetV2, ViT DeiT and Swin Transformer models when training from scratch and pretraining on the MedNIST medical dataset rather than the ImageNet dataset of natural images. The comparison is made by reporting six performance metrics, the Scott–Knott Effect Size Difference, Wilcoxon statistical test and the Borda Count method. We also use the Grad-CAM algorithm to study the model's interpretability. Finally, the model's robustness is tested by evaluating it on Gaussian noised images.

Although pretrained MobileNetV2 was the best model in terms of performance, the best model in terms of performance, interpretability, and robustness to noise is the trained from scratch Swin Transformer using the CXR (accuracy = 93.21 per cent) and CT (accuracy = 94.14 per cent) modalities.

Models compared are pretrained on MedNIST and leverage both the CT and CXR modalities.

Recognising the as-built state of construction elements is crucial for construction progress monitoring. Construction scholars have used computer vision-based algorithms to automate this process. Robust object recognition from indoor site images has been inhibited by technical challenges related to indoor objects, lighting conditions and camera positioning. Compared with traditional machine learning algorithms, one-stage detector deep learning (DL) algorithms can prioritise the inference speed, enable real-time accurate object detection and classification. This study aims to present a DL-based approach to facilitate the as-built state recognition of indoor construction works.

The one-stage DL-based approach was built upon YOLO version 4 (YOLOv4) algorithm using transfer learning with few hyperparameters customised and trained in the Google Colab virtual machine. The process of framing, insulation and drywall installation of indoor partitions was selected as the as-built scenario. For training, images were captured from two indoor sites with publicly available online images.

The DL model reported a best-trained weight with a mean average precision of 92% and an average loss of 0.83. Compared to previous studies, the automation level of this study is high due to the use of fixed time-lapse cameras for data collection and zero manual intervention from the pre-processing algorithms to enhance visual quality of indoor images.

This study extends the application of DL models for recognising as-built state of indoor construction works upon providing training images. Presenting a workflow on training DL models in a virtual machine platform by reducing the computational complexities associated with DL models is also materialised.

An abundance of techniques has been presented so forth for waste classification but, they deliver inefficient results with low accuracy. Their achievement on various repositories is different and also, there is insufficiency of high-scale databases for training. The purpose of the study is to provide high security.

In this research, optimization-assisted federated learning (FL) is introduced for thermoplastic waste segregation and classification. The deep learning (DL) network trained by Archimedes Henry gas solubility optimization (AHGSO) is used for the classification of plastic and resin types. The deep quantum neural networks (DQNN) is used for first-level classification and the deep max-out network (DMN) is employed for second-level classification. This developed AHGSO is obtained by blending the features of Archimedes optimization algorithm (AOA) and Henry gas solubility optimization (HGSO). The entities included in this approach are nodes and servers. Local training is carried out depending on local data and updations to the server are performed. Then, the model is aggregated at the server. Thereafter, each node downloads the global model and the update training is executed depending on the downloaded global and the local model till it achieves the satisfied condition. Finally, local update and aggregation at the server is altered based on the average method. The Data tag suite (DATS_2022) dataset is used for multilevel thermoplastic waste segregation and classification.

By using the DQNN in first-level classification the designed optimization-assisted FL has gained an accuracy of 0.930, mean average precision (MAP) of 0.933, false positive rate (FPR) of 0.213, loss function of 0.211, mean square error (MSE) of 0.328 and root mean square error (RMSE) of 0.572. In the second level classification, by using DMN the accuracy, MAP, FPR, loss function, MSE and RMSE are 0.932, 0.935, 0.093, 0.068, 0.303 and 0.551.

The multilevel thermoplastic waste segregation and classification using the proposed model is accurate and improves the effectiveness of the classification.

Higher energy conversion efficiency of internal combustion engine can be achieved with optimal control of unsteady in-cylinder flow fields inside a direct-injection (DI) engine. However, it remains a daunting task to predict the nonlinear and transient in-cylinder flow motion because they are highly complex which change both in space and time. Recently, machine learning methods have demonstrated great promises to infer relatively simple temporal flow field development. This paper aims to feature a physics-guided machine learning approach to realize high accuracy and generalization prediction for complex swirl-induced flow field motions.

To achieve high-fidelity time-series prediction of unsteady engine flow fields, this work features an automated machine learning framework with the following objectives: (1) The spatiotemporal physical constraint of the flow field structure is transferred to machine learning structure. (2) The ML inputs and targets are efficiently designed that ensure high model convergence with limited sets of experiments. (3) The prediction results are optimized by ensemble learning mechanism within the automated machine learning framework.

The proposed data-driven framework is proven effective in different time periods and different extent of unsteadiness of the flow dynamics, and the predicted flow fields are highly similar to the target field under various complex flow patterns. Among the described framework designs, the utilization of spatial flow field structure is the featured improvement to the time-series flow field prediction process.

The proposed flow field prediction framework could be generalized to different crank angle periods, cycles and swirl ratio conditions, which could greatly promote real-time flow control and reduce experiments on in-cylinder flow field measurement and diagnostics.

In this digital era, email is the most pervasive form of communication between people. Many users become a victim of spam emails and their data have been exposed.

Researchers contribute to solving this problem by a focus on advanced machine learning algorithms and improved models for detecting spam emails but there is still a gap in features. To achieve good results, features also play an important role. To evaluate the performance of applied classifiers, 10-fold cross-validation is used.

The results approve that the spam emails are correctly classified with the accuracy of 98.00% for the Support Vector Machine and 98.06% for the Artificial Neural Network as compared to other applied machine learning classifiers.

In this paper, Point-Biserial correlation is applied to each feature concerning the class label of the University of California Irvine (UCI) spambase email dataset to select the best features. Extensive experiments are conducted on selected features by training the different classifiers.

The Chinese housing market has gone through rapid growth during the past decade, and house price forecasting has evolved to be a significant issue that draws enormous attention from investors, policy makers and researchers. This study investigates neural networks for composite property price index forecasting from ten major Chinese cities for the period of July 2005–April 2021.

The goal is to build simple and accurate neural network models that contribute to pure technical forecasts of composite property prices. To facilitate the analysis, the authors consider different model settings across algorithms, delays, hidden neurons and data spitting ratios.

The authors arrive at a pretty simple neural network with six delays and three hidden neurons, which generates rather stable performance of average relative root mean square errors across the ten cities below 1% for the training, validation and testing phases.

Results here could be utilized on a standalone basis or combined with fundamental forecasts to help form perspectives of composite property price trends and conduct policy analysis.