Abstract Sentiment analysis is a technique for assessing people's opinions and viewpoints based on the text or photos they publish on social networking sites such as Instagram and Twitter. Because it can be difficult to pinpoint the precise thoughts, ideas, and sentiments expressed in a text document or image, sentiment classification is a difficult undertaking. People share their emotions in diverse ways depending on the situation and subject. In this paper, we proposed a hybrid feature extraction and selection technique using numerous deep learning techniques. The features extracted from the image, such as luminance, chrominance, histogram, autoencoder, etc., are validated with a modified convolutional neural network called mCNN. The number of deep CNN layers, size of extracted features, various activation functions, and different optimizers have been used for CNN feeding. In an extensive experimental analysis, our module was tested and compared with two different deep learning modules, such as RESNET and VGGNET. Our proposed module obtains higher accuracy than two conventional deep learning frameworks.
Alan : Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|