A Unified AI-powered Social Media Platform for Intelligent Scheduling and Data Driven Analytics Using Multi-Layered Artificial Neural Networks (ANNs)

https://doi.org/10.5281/zenodo.17572988

Abstract

The paper presents a new process of data analytics of users of social networks (SN) based on the use of artificial neural networks (ANN). , which allows engaging with consumers in real-time, Data sentiment analysis. Nevertheless, the existing Deep learning models deal with single-task aspects and do not dynamically model user interactions and sentiment changes. In this regard, we put forward an AI combined Unified Hybrid Attention-based Personalized Artificial Deep Neural Network (A-DNN), a single deep learning system that incorporates hierarchical attention, hybrid optimization, and personalized brand suggestions. A-DNN is an adaptive, dynamically prioritized engagement signal-based and sentiment signal-based hyper-parameter tuning with Moth-Flame Optimization (MFO) that enhances training efficiency and offers a 95% accuracy to the proposed model on the training set and a 92% accuracy to the proposed model on a testing set. Massive experiments show that A-DNN is more effective in engagement prediction Data with an A-DNN 17.5% minimized MSE, sentiment classification accuracy Data with a 4.1 percentage increase, and prediction performance compared to the existing methods. The trend in the analysis and application of social media currently is the use of pre-determined features, which are dedicated to the subsequent modules of development of the model, to fulfill the final tasks. Representation learning has been a central issue in machine learning and has been generally accepted as a crucial issue in the performance of end tasks. This paper presents evidence that there will be specially learned features that will handle the heterogeneous, collective, and diverse nature of social media data. Hence, we suggest moving the emphasis from model development to latent feature learning and introducing a coherent system of latent feature learning in social media. Popular deep learning is used to overcome the noisy, diverse, heterogeneous, and interconnected properties of social media data because of its outstanding abstract capabilities. More specifically, we operationalize a new relational deep learning model to answer the social media link analysis BY creating a multimodal deep learning model towards the social image retrieval problem. We demonstrate the fact that the resulting latent features result in the improvement of both social media tasks.