The toxic comment detection over the internet through social networking posts found hatred comments and apply certain limitations to stop the negative impact of that information in our society. In order to perform sentiment analysis, NLP text classification approach is very effective. In this paper, we design a specific algorithm using Convolution Neural Network (CNN) approach and perform TextBlob sentiment analysis to evaluate the polarity and subjectivity analysis of posted tweets or comments. This paper can also filter the tweets collected over different locations formed Twitter dataset and then model is evaluated in terms of accuracy, precision, recall and f1-score as calculated results of 0.984, 0.887, 0.905 and 0.895 respectively for the analysis of toxic/non-toxic comment identification. Hence, our algorithm utilized NLTK and TextBlob libraries and suggests that the analyzed post can be recommended to the others or not.