ANALYSIS OF TWITTER USERS' SENTIMENT BASED ON TEXT DATA USING CONVOLUTIONAL NEURAL NETWORK

The article considers using convolutional neural networks (CNN) to analyze user sentiment on the Twitter platform based on text data. The relevance of the work is due to the growing amount of unstructured text data in social networks and the need for their effective processing to understand public opinion. The proposed CNN model includes representing text data in vector space, convolutional and pooling layers for extracting relevant features, and fully connected layers for classifying sentiment into positive and negative. To prevent overfitting, we applied the Dropout layers to exclude neurons randomly. An experimental evaluation of the model was conducted on a specially prepared dataset from Twitter. The test results demonstrate the prospects of using CNN for automated analysis of social media users' sentiments.