Below the dotted line is the processing process of real-time calculation in the room, after the user posts and replies, write it to the Kafka message queue, first of all, the whitelist user filter, and then one is based on the training effect of the classification model, the text is identified and classified, the spam and non-junk posts are distinguished; Recently, Mei grapefruit to the algorithm made a further iteration, the user's nickname, posting interval, posting behavior analysis, because of the use of anti-spam algorithm, so that mei grapefruit APP well prevent spam.