Systems and Soft Computing最新文献

In today's digital era, social media has become a new tool for communication and sharing information, with the availability of high-speed internet it tends to reach the masses much faster. Lack of regulations and ethics have made advancement in the proliferation of abusive language and hate speech has become a growing concern on social media platforms in the form of posts, replies, and comments towards individuals, groups, religions, and communities. However, the process of classification of hate speech manually on online platforms is cumbersome and impractical due to the excessive amount of data being generated. Therefore, it is crucial to automatically filter online content to identify and eliminate hate speech from social media. Widely spoken resource-rich languages like English have driven the research and achieved the desired result due to the accessibility of large corpora, annotated datasets, and tools. Resource-constrained languages are not able to achieve the benefits of advancement due to a lack of data corpus and annotated datasets. India has diverse languages that change with demographics and languages that have limited data availability and semantic differences. Telugu is one of the low-resource Dravidian languages spoken in the southern part of India.

In this paper, we present a monolingual Telugu corpus consisting of tweets posted on Twitter annotated with hate and non-hate labels and experiments to provide a comparison of state-of-the-art fine-tuned deep learning models (mBERT, DistilBERT, IndicBERT, NLLB, Muril, RNN+LSTM, XLM-RoBERTa, and Indic-Bart). Through transfer learning and hyperparameter tuning, the models are compared for their effectiveness in classifying hate speech in Telugu text. The fine-tuned mBERT model outperformed all other fine-tuned models achieving an accuracy of 98.2. The authors also propose a deployment model for social media accounts.

As the growth of internet technology, human life is full of various advertisements. It is possible for individuals to obtain the advertising information they require, whether in an online or offline context. A research proposal is presented with the objective of enhancing the precision of online advertising recommendations. The proposal is based on the design of internet search engine advertising text keyword recommendation models, which integrate entity naming recognition models to facilitate tasks such as text classification and feature extraction. A recommendation algorithm based on content similarity is used to achieve keyword recommendation. Under the similarity calculation method of continuous bag-of-words model, when K is 100, the model weighted precision of the feature extraction method based on graph sorting and inverse text frequency index is 0.88, the weighted recall is 0.76, and the weighted F1-score is 0.82. In offline simulation testing, 85 % of the keyword recommendation model's recommendation time is less than 1 s, 99 % of the recommendation time is less than 2 s, and the recommendation cost can be significantly reduced by 75 %. In practical applications, the recommendation efficiency of this method can reach 96.3 %, and the recommendation precision can reach 95.8 %. The recommended satisfaction rate can reach 99.5 %. The results demonstrate that this method can provide highly accurate keyword recommendations and reduce the cost of advertising placement. Furthermore, it has been recognized and praised by users.

Image fusion is a popular research direction in the field of computer vision. Traditional image fusion algorithms can achieve good results in fusing grayscale images, but it is difficult to achieve ideal results in processing multi-spectral images. To address the shortcomings of multi-spectral image fusion, this study proposes a low computational complexity and low latency multi-spectral image fusion model by utilizing a multi-step degree moment matching algorithm and a generative adversarial network for fusion. Through experiments, it was found that the F1 score of the GAN-MMN model on the TinyPerson dataset was 89.79 %, with an average recall rate of 89.76 %. The GAN-MMN performance was higher than that of the control model. Meanwhile, the GAN-MMN model also exhibited superior performance in high-frequency feature extraction and time delay compared to the control model. According to the experimental results, the proposed multi-spectral remote sensing image fusion model had a high feature extraction effect, and its recall rate and F1 score were better than the control model, so the research model had a certain progressiveness. The proposal of this model gives a new approach for the processing of multi-spectral remote sensing images, effectively promoting the development of the computer vision industry.

Video analysis for violence detection is crucial, especially when dealing with crowd data, where the potential for severe mob attacks in sensitive areas is high. This paper proposes a solution utilizing Convolutional Restricted Boltzmann Machine (CRBM) for video analysis, integrating the strengths of Convolutional Neural Network (CNN) and Restricted Boltzmann Machine (RBM). By focusing on image patches rather than entire frames, the method addresses the challenge of object detection in crowded scenes. The CRBM combines deep-level image analysis from CNN with unsupervised feature extraction in RBM, facilitated by image convolution using Gabor filters in the hidden layer. Dropout regularization mitigates overfitting, enhancing model generality. Extracted features are inputted into an SVM classifier for face detection and a custom VGG16 model for emotion identification. Event probability is then determined through logistic regression based on facial expressions. Despite existing approaches for smart crowd behaviour identification, there remains a tradeoff between accuracy and processing time. Our proposed solution addresses this by employing proper frame preprocessing techniques for feature extraction. Validation using quantitative and qualitative metrics confirms the effectiveness of the approach.

With the advancement of semiconductor technology, pedestrian navigation and positioning technology based on smartphones is becoming increasingly important in people's travel. However, precise positioning is challenging due to the use of inertial measurement units in low-cost smartphones and the complex motion states of pedestrians. To navigate and locate pedestrians in complex motion states, a method for converting between smartphone coordinate systems and navigation coordinate systems was studied and designed, and the errors of the built-in sensors of smartphones were analyzed and calibrated. In addition, support vector machines were used to optimize pedestrian trajectory prediction algorithms, and a pedestrian motion state recognition algorithm was designed based on this. To solve the classification problem of multiple human motion states, a multi classification model was constructed and adjacent gait correlation constraints were introduced to correct the classification results. Experiments indicated that the sum of squared errors for traditional algorithms estimating pedestrian trajectories was 0.92, whereas the optimized algorithms produced an improved sum of squared errors of 0.26. Consequently, the average sum of squared errors was reduced by 71.74 %, and the convergence speed increased by 55.56 %. The pedestrian trajectory prediction algorithm optimized by support vector machine could significantly lift the positioning and navigation efficiency, with a correct recognition rate of over 93 % and a position recognition accuracy of 78.8 % - 88.4 %. By optimizing recognition of the motion state of pedestrians, more accurate determination of their position and motion state can be achieved.

This paper presents an optimal vehicle routing model for an efficient waste collection process using the Ogun State Waste Management Agency (OGWAMA) as a case study. Just like in many cases, the current manual predetermined routing method used by OGWAMA is inefficient and contributes to excessive fuel usage. These challenges, in addition to the small instances reported in most literature, inspire this research to propose an improved routing scheme that takes into account real-time costs and eventually develops a novel instance based on OGWAMA's operation mode. The developed model was optimized using a new discrete smell agent optimization (SAO) algorithm and compared to firefly algorithm (FA) and particle swarm optimization (PSO). The SAO outperformed FA and PSO, achieving 3.92 % and 19.38 % improvements in service cost (SC) and 2.65 % and 14.96 % improvements in total travel distance (TTD), respectively. The convergence rates of the algorithms were also compared; using the Optimized Depot (OD) techniques and results shows the acceptability of the proposed approaches.

This paper presents a comparative analysis between two state-of-the-art object detection models, DETR and YOLOv8, focusing on their effectiveness in fruit detection for yield prediction in agriculture. The study begins with data acquisition, utilizing images and corresponding annotations to train and evaluate the models. Our approach employs a data-driven methodology, dividing the dataset into training and testing sets, with rigorous validation to ensure robustness.

For DETR, evaluation results demonstrate promising performance across various IoU thresholds, indicating its effectiveness in accurately localizing fruits within bounding boxes. Additionally, YOLOv8 exhibits substantial improvements in detection performance, achieving high precision and recall rates, particularly noteworthy for "orange" and "sweet_orange" classes. Notably, the model showcases commendable proficiency even in challenging scenarios.

In conclusion, both DETR and YOLOv8 offer valuable insights for precision farming, aiding farmers in yield prediction and harvest planning. While DETR demonstrates robustness and efficiency in fruit detection, YOLOv8 excels in high-precision detection, albeit with longer training times. These findings highlight the potential of advanced object detection models in revolutionizing agricultural practices, contributing to enhanced productivity and market equilibrium.

Fingerprints are considered as one of the most important and prominent feature for an individual identification. Due to their consistency and reliability in biometric feature identification, they are most widely used for biometric recognition systems. In these systems, the relevant feature extraction plays important role in achieving required classification accuracy. In recent time, deep learning techniques are being used for fingerprint recognition with more accuracy and efficient results. Major difficulty which has been reported in previous researches, is the limited size of samples. Therefore, we propose two approaches, inversion and multi augmentation to augment the sample size with newly generated images for each feature map. Besides this, multiple networks are used simultaneously for feature extraction from newly generated images in parallel mode. Deep neural network architectures are used with proposed inversion methods and multi augmentation methods to classify the samples of fingerprints for personnel identification and verification. Pre-trained deep convolutional models like VGG16, VGG19, ResNet50 and InceptionV3 are fine-tuned with new processed fingerprint images for feature extraction and classification. The collective samples of fingerprints have been classified into 10 classes. The simulation results have been obtained with different optimizers and it has been observed that VGG 19 model exhibits the accuracies of 88 % and 93 % with inversion and multi augmentation approaches respectively. Whereas, VGG16 model exhibits 93 % with inversion approach and 97 % with multi augmentation approach. Thus, the proposed approach exhibits the accuracy up to 97 % with VGG16 model which is significantly much higher than that of any other model with the same dataset FVC2000_DB4.

Users are increasingly turning to big data processing systems to extract valuable information from massive datasets as the field of big data grows. Data analytics platforms are used by e-commerce enterprises to improve product suggestions and model business processes. In order to meet the needs of large-scale data center operation and maintenance management, Internet companies often use Flink to process log data. This paper takes the big data processing and analysis platforms built by Internet financial companies and large banks as examples, and implants a stock prediction model based on Deep Neural Network (DNN). In this context, this paper completes the following work: 1) The research status of big data processing and analysis platforms at home and abroad is introduced. 2) Drawing on the modular design idea, the commercial bank big data platform is designed and the functions of each sub-module are introduced. Then the basic principle and structure of Convolutional Neural Networks (CNN) are expounded. 3) The optimal parameters of Convolutional Neural Networks are selected through experiments, and then the trained model is used for experiments. It can be seen that the stock prediction model proposed in this article has a higher prediction accuracy compared to existing models, which also verifies the validity of the proposed model. Input the data and compare the obtained results with the actual results, and finally show that the model in this paper has a good performance on stock prediction.

With the development of the economy, more and more electronic manufacturing enterprises are emerging like mushrooms after rain. These enterprises, while developing, also face financial risks caused by various reasons. In order to provide early warning for financial risks of enterprises, improve the accuracy of identifying financial risks, avoid financial crises, and provide assistance for sustainable development decisions, this paper proposes a financial management model based on modified random forest. In order to improve the generalization ability of financial management models, pruning methods were adopted in the study to avoid overfitting. Synthetic minority oversampling technique is used to optimize the financial management model and reduce the calculation deviation of the model through its sampling ability. At the same time, the prediction index system is proposed to improve the analysis ability of the financial management model. The results show that the accuracy and recall rate of the improved algorithm based on random forest proposed in this study in identifying corporate financial distress are 98.03 % and 100 % respectively. The importance value of operating income and cash flow in enterprise indicators is 0.391, which is the most relevant indicator for enterprise financial forecasting. The results show that after the improvement of synthetic minority oversampling technique, the stochastic forest model can effectively improve the recognition and early warning ability of enterprises’ financial distress, and is conducive to maintaining good operating efficiency and sustainable operation of enterprises. Electronic manufacturing enterprises need to strengthen their attention to cash flow, improve their cash flow, and enhance their profitability. The financial management model designed by the research institute can provide technical and information support for financial early warning and sustainable development of electronic manufacturing enterprises.