@article {10.3844/jcssp.2026.2367.2384,
article_type = {journal},
title = {A Sentiment Analysis Approach for Affective State Recognition With SWISH-Activated DenseNet-169 in Human Event Forecasting},
author = {Elangovan, Karthik and Paulraj, Prabhakaran and G L, Dinesh Babu and  Anakath, A. S. and Ramasamy, Jayaraj},
volume = {22},
number = {8},
year = {2026},
month = {Jul},
pages = {2367-2384},
doi = {10.3844/jcssp.2026.2367.2384},
url = {https://thescipub.com/abstract/jcssp.2026.2367.2384},
abstract = {Detecting human feelings from images is perhaps one of the most useful and challenging social communication problems in research. Emotion detection-based deep computing outperforms classic image processing techniques. Facial recognition technology has permeated every aspect of our daily lives, from unlocking our phones to getting private file access on our systems. Machine Learning and Deep Learning make it easy to detect facial expression by using the method of sentiment analysis, which in turn makes it possible to identify the status of the emotions, which can be used to predict human emotions. ML has opened a new world of utilizing technology, from self-driving automobiles to recognizing faces on your mobile lock screen. Face recognition finds individuals in the landscape using AI, ML and DL algorithms. Further confirmations using big datasets with both positive and negative photographs, after all the facial traits have been recorded, aid in confirming that the image is truly of a human face. Face detection, feature extraction, and emotion categorization are all part of the basic facial emotion identification process. The user’s facial expression is dynamically captured from a video stream or statically captured image, and the image is further processed. The seven human facial emotions are classified in the processed image by applying machine learning-based proposed SWISH DenseNet-169 image classifier thereby it facilitates the forecasting of human moves. In this work, SWISH DenseNet-169 is contrasted with cutting-edge techniques such as EfficientNet, CNN, Vision Transformers, RCNN, and LSTM. The proposed DenseNet-169 is enhanced the accuracy measures and it was executed on a variety of hyman expression data sets includes CK+, JAFFF, BES, EMOTI-W, and IAP.},
journal = {Journal of Computer Science},
publisher = {Science Publications}
}