International Journal on Recent and Innovation Trends in Computing and Communication
Not a member yet
8613 research outputs found
Sort by
Automating Data Labeling and Annotation Pipelines for Large Language Models (LLMs) in the Financial Industry using Machine Learning
Get PDFThe growing magnitude and intricacy of financial information present considerable obstacles for machine learning systems, especially Large Language Models (LLMs), which necessitate extensive, high-caliber labeled datasets for training. Conventional manual labeling approaches are ineffective and expensive, constraining the expandability of LLMs in the finance sector. This research introduces an automated data labeling and annotation framework utilizing Principal Component Analysis (PCA) and Decision Trees (DT), two robust machine learning methodologies, to optimize and improve the labeling procedure for financial information. PCA is utilized for reducing dimensionality, assisting in the identification of crucial features and trends in financial datasets, while DTs are employed to categorize data and automate the annotation process. The proposed system aims to enhance the precision, effectiveness, and scalability of the data labeling procedure, ultimately facilitating the ongoing training of LLMs with contextually pertinent, labeled financial data
Impact of Feature Extraction Combined with Data Sampling Methods on Heartbeat Categorization
Get PDFDealing with class-imbalanced datasets in data analytics poses challenges, especially when faced with high-dimensional data. In order to handle this issue, researchers often utilize preprocessed methods like feature selection. Feature selection attempts to create a more informative and condensed feature set, while data sampling helps alleviate class imbalance. In our study, aim is to explore the effectiveness of data sampling preprocessed techniques combined with feature extraction using a dataset on ECG Heartbeat. We evaluate ensemble classifiers: Decision Tree; Random Forests (RF), Gradient-Boosted Trees (GBT) for feature extraction. In terms of data sampling, we assess the effectiveness of two methods: Random Under sampling (RUS) and Synthetic Minority Oversampling (SMOTE). The performance of this feature extraction is measured using the sensitivity and the specificity, two important metrics used for accuracy. Our findings depict that the combination of the RUS and GBT method yields the highest performance for ECG Heartbeat detection
Integrating Local Binary Pattern Image Transformations and Customized Deep Learning Models for Enhanced Fetal Cardiac Anomaly Detection
Get PDFThis research focuses on developing a deep learning diagnostic model for diagnosing fetal cardiac anomalies from real time ultrasound scan images. The dataset framed in the previous research is transformed using Local Binary Pattern (LBP) technique which is trained with the deep learning models to create the classifiers. These models are used to classify the Real Time images captured by ultrasound scanning machines. The LBP is a texture image feature. LBP captures the local structure and patterns within an image by comparing the intensity values of each pixel with its surrounding neighbours. The LBP operator assigns a binary code to each pixel based on the comparison results, resulting in a texture representation of the image. The FetaEcho_V05 dataset is transformed into an LBP image dataset titled as FetalEcho_V0501. This dataset is used for creating the classifiers by training the custom CNN(CCNN), AlexNet, VGG16 and ResNet50 deep learning models. The classifiers are evaluated for its overall classification performance and class wise evaluation performance using the metrics the precision, recall, accuracy and F1 score metrics. When the overall performance is considered, the CCCNN model performed the best on the FetalEcho_V0501 dataset
How Good Is Local Search for Capacitated Facility Location Problem: An Experimental Study
Get PDFFacility location problems have been widely studied since 1960’s. These problems are known to be strongly NP-hard. In capacitated variant of the problem, a capacity constraint is associated with each facility. Capacitated facility location problem (CFLP) instances can be solved exactly using existing MILP solvers but only for small instance sizes. As the size of the problem instance increases beyond few hundred facilities and few hundred clients, it becomes prohibitive to solve these instances exactly. For large problem instances, therefore, other solution methods are used. One approach is to use heuristic methods. These methods usually give good solutions in reasonable time but they do not provide any guarantee about the quality of the solution. Somewhere between these two extremes exist another class of algorithms called approximation algorithms. They also provide only suboptimal solutions to the problem, like heuristic algorithms, in polynomial time. How ever they guarantee worst case upper bounds on the cost of the solution. So, a solution obtained using an approximation algorithm is guaranteed to have its cost between the optimal cost and the upper bound. We present experimental studies done with a local search based approximation algorithm for CFLP given by Bansal et al. [1] to show that this algorithm performs well in practice
Study the Automated System for Tracking the Movement Of Feeding Cows in the Cloud
Get PDFNowadays, in order to improve productivity and make sure that animals are healthy, it is crucial to use current technology in agriculture. A revolutionary technology that integrates the IoT, cloud, data analytics, & automated feeding cow tracking in the cloud is a game-changer for managing and overseeing cattle operations on a daily basis. Many tasks related to cattle raising are still done by hand on farms today. Specifically, most farms depend on the farmer's eye for animal health rather than on machinery. It is possible for managers to forecast the health of farm animals based on data collected from monitoring their behavior. We present a WSN-based livestock monitoring system in this article. With the use of internet of things (IoT) devices and cloud computing, the suggested system can keep tabs on livestock. The livestock's movements were tracked by attaching IoT collars on their necks. By uploading data from livestock observation systems to cloud platforms, farming managers can keep tabs on real-time data. We found out through testing that the suggested method can keep tabs on farm animals in real time
SnO2-ZnO Nanocomposite Generated Robust Conducting Polymers for Sensing
Get PDFThe exploration embarks on the alchemy of synthesizing, orchestrating self-assembly, and unraveling the enigmatic properties of ZnO nanostructures and nanocomposites. The chapters unfold like a poetic narrative, weaving tales of production, characterization, and the magical applications of zinc oxide nanoparticles. In the initial act, the spotlight falls on the stage of asymmetric ZnO nanostructures adorned with interior cavities. Here, the script unveils the drama of structurally anisotropic wonders, where newly formed inner spaces grace the top realms. Unlike their counterparts with hollow cores, enter surfactants as virtuoso conductors, choreographing the intricate dance of fundamental nano-crystallites and orchestrating the harmony of multiple crystal planes within ZnO's asymmetric nanostructures, as if whispered by the muses themselves. Enter the second act, where hydrothermal alchemy conjures hourglass-shaped ZnO nanostructures. The narrative unfolds, revealing the secret script of ZnO subunits and their delicate self-assembly ballet guided by the ethereal presence of Tween-85. The hourglass structures emerge like linear poetry, their verses composed through the enchanting van der Waals interactions between subunits' surface-anchored alkylated oleate groups. The revelation of van der Waals interactions on surfaces becomes the climax, an unforeseen twist in the narrative, unveiled as the hourglass structures gracefully unravel in a poetic disassembly
Deep Stacked CNN-LSTM (DS-CNN-LSTM) based Spectrum Sensing in Cognitive Radio
Get PDFThe multidimensionality of spectrum sensing, the intrinsic complexity of its dependence, and the unpredictability associated with spectrum data all contribute to the difficulty of the task. The network of cognitive radio (CR) is comprised of both primary and secondary users inside its network. The SUs that are part of the CR network are able to identify the spectrum band and access white space in an opportunistic manner. Enhancing spectrum efficiency may be accomplished by using white spaces. This study presents a Deep Stacked CNN-LSTM (DS-CNN-LSTM)-based spectrum sensing strategy that learns implicit features from spectrum data, such as temporal correlation. This approach is based on the research that we have conducted. The effectiveness of the recommended method is shown by a sufficient number of simulations, and the results of the simulations demonstrate that it outperforms the current state of the art in terms of detection probability and classification accuracy. A comparison is made between the most cutting-edge spectrum sensing approaches and the DS-CNN-LSTM method that has been recommended. The results of the experiments indicate that the proposed methods improve detection performance and classification accuracy even when the signal-to-noise ratio is low. As we can see, the improvement that was achieved comes at the price of a longer amount of time spent on training and a little increase in the amount of time spent on execution
Decision-Tree-based Ensemble Learning Models for Long-Term Traffic Intensity Forecasting and Analysis of Congestion Treatment Strategies
Get PDFTraffic intensity forecasting is a key factor in analyzing traffic patterns and making recommendations to overcome congestion. It can also prove helpful to the Intelligent Transportation System (ITS) application. In this work, we have made a detailed comparative evaluation of various ML regression algorithms aimed at solving the problem of long-term traffic intensity prediction. A lot of work focuses mainly on traffic flow prediction. However, work on traffic intensity prediction has not been done sufficiently. For this problem, ensemble learning methods like Random Forest Regression that use the outputs of individual trees (Decision Trees) proved to be more successful and efficient rather than the single models approach. This work also dictates the study of various features that may be used to express the traffic data and the various strategies that can be employed to make decisions on whether a solution to overcome traffic congestion is needed
An Intelligent Multimodal Emotion Recognition System for E-Learning
Get PDFThe purpose of this research paper is to introduce an Intelligent Multimodal Emotion Recognition System (IMERS) that aims to improve the e-learning process by accurately perceiving and reacting to the emotional states of learners. IMERS incorporates information through three primary modalities: facial expressions, voice, and text. The utilization of multimodal fusion in this technique effectively addresses the constraints of single-modality systems and yields a more extensive and precise comprehension of emotions. The paper highlight the following: Initially, we discuss the structure of multimodal fusion, specifically focusing on the architecture of IMERS. This includes an explanation of the many components involved, such as data preparation, feature extraction, decision fusion, and sentiment classification. Every approach employs distinct deep learning algorithms customized to its unique properties. Furthermore, our assessment of IMERS encompasses its proficiency in discerning emotions inside e-learning environments, as evidenced by its correct detection of primary emotions across diverse datasets. Another area of emphasis is personalized learning applications, in which we demonstrate how IMERS customize learning experiences by adapting instruction, offering specific feedback, and cultivating an emotionally nurturing learning environment.
A Multimedia Cloud Computing Model for Combinatorial Virtual Machine Placement
Get PDFCloud computing, which allows users to access subscription-based services on a pay-as-you-go basis, has recently transformed IT departments. Today, a variety of media services are offered through the Internet owing to the development of multimedia cloud computing, which is based on cloud computing. However, as multimedia cloud computing spreads, it has a negative influence on greenhouse gas emissions due to its high energy consumption and raises expenses for cloud users. Therefore, while still providing consumers with the resources they require and maintaining a high level of service, multimedia cloud service providers should make every effort to consume as little energy as possible. This proposal proposes residual usage-aware (RUA) and performance-aware (PA) methods for virtual machine placement. To save energy, find a suitable host to switch off. These two techniques were merged and applied to cloud data centers in order to complete the VM consolidation process. The outcomes of the simulation demonstrate a trade-off between energy consumption and SLA violations. Additionally, during VM deployment, it can manage shifting workloads to prevent host overload, dramatically lowering SLA breaches