8333 research outputs found
Sort by
Sentiment Evaluation of FlipKart Product Reviews using the Recurring Neural Network
No full textSentiment analysis, a key aspect of natural language processing, is critical in deciphering user emotions from textual data. This paper presents a comprehensive exploration of sentiment analysis, focusing on the development and optimization of a deep learning model. Leveraging LSTM networks, our study involves training the model on diverse datasets, encompassing a range of sentiments. We delve into preprocessing techniques and feature engineering to enhance model robustness. Results showcase the model's effectiveness in classifying sentiments, with a particular emphasis on practical applications such as customer reviews and social media comments. The achieved accuracy, precision, and recall metrics demonstrate the model's potential for real-world implementations
Prediction of Resource Utilization in Cloud Computing using Machine Learning
No full textIn today’s modern computing infrastructure, cloud computing has emerged as a pivotal paradigm, offering scalability and flexibility to satisfy the demands of a wide variety of specific applications. Maintaining optimal performance and cost-effectiveness inside cloud settings continues to be a significant problem, and one of the most important challenges is efficient resource utilisation. A resource utilisation prediction system is required to aid the resource allocator in providing optimal resource allocation. Accurate prediction is difficult in such a dynamic resource utilisation. The applications of machine learning techniques are the primary emphasis of this research project, which aims to predict resource utilisation in cloud computing systems. The dataset GWA-T-12 bitbrains (from distributed datacenter) have provided the data of timestamp, cpu usage, network transmitted throughput and Microsoft Azure traces has provided the data of cpu usage of cloud server. To predict VM workloads based on CPU utilisation, we use machine learning models such as Linear Regression, Decision Tree Regression, Gradient Boosting Regression, and Support Vector Regression, as well as deep learning architectures such as Long Short-Term Memory (LSTM) and Bi-directional Long Short-Term Memory (BiLSTM). The Python programming language is used to carry out the implementation within the Google Colab environment. Bi-directional Long Short Term Memory approach is considered more effective as compared to other models in terms of CPU Utilisation and Network Transmitted Throughput as it R2 score is close to 1 hence can produce more accurate results
Enhancing Cloud Security through Efficient Polynomial Approximations for Homomorphic Evaluation of Neural Network Activation Functions
No full textCurrent security cloud practices can successfully protect stored data and data in transit, but they do not keep the same protection during data processing. The data value extraction requires decryption, creating critical exposure points. As a result, privacy-preserving techniques are emerging as a crucial consideration in cloud computing. The homomorphic processing of machine learning models in the cloud represents a central challenge. The activation function is fundamental in constructing a privacy-preserving Neural Network (NN) with Homomorphic Encryption (HE). Standard activation functions require operations not supported by HE, so it is necessary to find cryptographically compatible replacement functions to operate over encrypted data. Multiple approaches address the limitation of function compatibility with polynomial approximation. These functions should exhibit a trade-off between complexity and accuracy, limiting the efficiency of conventional approximation techniques. The current literature on polynomial approximation of NN activation functions still lacks a thorough review. In this paper, we comprehensively review the standard activation functions of modern NN models and current polynomial approximation approaches. We highlight fundamental features to consider in the activation function and the approximation technique to operate over encrypted data
Enhancing Cloud Security: Implementing and Evaluating the Zero Trust Architecture with Firebase Services and Advanced Encryption Algorithms
No full textTo enhance the cloud security developing a web-based user management system integrating authentication while implementing the Zero-trust model. The implementation leveraged tools like Firebase for authentication and data handling, hosted on AWS servers. The project addressed prevalent concerns in modern cybersecurity, emphasizing the need for robust access control and data security in web applications. This model assumes no implicit trust, validating every request explicitly and enforcing the principle of least privilege. The project's primary contribution lay in its holistic approach to user management, incorporating Firebase tools for authentication, custom claims, and real-time database handling. The system effectively validated users through Firebase authentication and secured data handling via Firebase's real-time database with meticulous security rules. The findings align with the current state of the art in cybersecurity by showcasing an application of the Zero-trust model within a web-based user management system. Cloud storage is becoming a dependable method of storing data, there are still significant security concerns that affect both cloud computing and cloud storage, including maintaining confidentiality integrity as well as confidentiality. The integration of advanced encryption algorithms, namely Double AES (DAES) and Blowfish, within cloud storage systems, particularly focusing on Amazon S3, to bolster data security. The research examines the implementation of these robust encryption methods and evaluates their impact on data transfer efficiency and system performance. The investigation involves encrypting files using DAES and Blowfish before uploading them to an Amazon S3 bucket. Through empirical analysis, this study assesses the encryption time, throughput, and overall performance implications of integrating these encryption algorithms into cloud storage
Traffic Flow Forecasting using DeepAR
No full textImproved traffic volume forecasting techniques are required due to disruptions in to travel patterns caused by the COVID-19 pandemic. During pandemics, traditional approaches find it difficult to handle complicated dynamics. This work investigates the use of an RNN architecture called DeepAR to forecast traffic volume both during and after the epidemic. DeepAR is a good fit for simulating pandemic-induced traffic patterns because of its capacity to manage temporal dependencies and long-range interactions. DeepAR models were trained for various prediction horizons using historical data covering pre-pandemic, pandemic, and post-pandemic eras. The results show that DeepAR works better than conventional techniques at capturing pandemic-induced changes in traffic patterns. Its flexibility makes proactive traffic control techniques possible. The implications of these findings for traffic management after the pandemic are noteworthy. Transportation authorities can improve traffic flow, manage infrastructure, and lessen congestion by utilising DeepAR. Because of its adaptability, DeepAR is a useful tool for transportation systems to adjust to constantly shifting traffic patterns
The Impact of a Befriending Service on Health-related Quality of Life in Older Adults: An Interventional N-of-1 Pilot Study (Preprint)
No full textPurpose: Befriending interventions are unlikely to reduce loneliness, but they may provide social support which buffers the negative impact of loneliness on health outcomes of older adults. An interventional N-of-1 design was used to assess the impact of a befriending intervention on health-related quality of life (HR-QoL) among older adults, and whether such intervention attenuated the impact of loneliness on HR-QoL.
Methods: Participants were n = 33 new users of the service, aged 60+. Outcomes were measured at 13 timepoints across 26 weeks, and data were analysed using generalised additive modelling (GAM) with a subset of data analysed using visual analysis.
Results: Results indicate that the befriending service may reduce decline of HR-QoL (i.e., health declined in the baseline phase over time: edf = 3.893, F = 3.0, p=0.002, while in the treatment phase, health remained more stable: edf = 5.98, F = 2.98, p=.008). The befriending intervention also suppressed the association between loneliness and HR-QoL.
Conclusion: We supported our hypothesis, that befriending interventions may moderate the impact of loneliness on HR-QoL. Interventional N-of-1 designs however carry considerable recruitment and participant burden, which should be considered prior to onset. This research provides an insight into practical difficulties when evaluating existing community-based services, particularly in relation to adhering to best practice design guidelines