There are various intense forces causing customers to use evaluated data when using social media platforms and microblogging sites. Today, customers throughout the world share their points of view on all kinds of topics through these sources. The massive volume of data created by these customers makes it impossible to analyze such data manually. Therefore, an efficient and intelligent method for evaluating social media data and their divergence needs to be developed. Today, various types of equipment and techniques are available for automatically estimating the classification of sentiments. Sentiment analysis involves determining people's emotions using facial expressions. Sentiment analysis can be performed for any individual based on specific incidents. The present study describes the analysis of an image dataset using CNNs with PCA intended to detect people's sentiments (specifically, whether a person is happy or sad). This process is optimized using a genetic algorithm to get better results. Further, a comparative analysis has been conducted between the different models generated by changing the mutation factor, performing batch normalization, and applying feature reduction using PCA. These steps are carried out across five experiments using the Kaggle dataset. The maximum accuracy obtained is 96.984%, which is associated with the Happy and Sad sentiments.

Sentiment analysis; convolutional neural networks; facial expression; genetic algorithm