Anti-social behaviour analysis using random forest and word to vector approach

Social Networking and micro-blogging applications provide active platforms for communications, sharing thoughts and ideas. Processing natural text coming from varied social platforms possess many technical challenges such as processing messages written in slang, informal short messages, classifying messages into different labels and category based on the meaning. Maximum natural text processing and interpretation systems use n-gram language models, which can be simple and powerful most of the time. Random forest ensemble-based classifier has the potential to generalise the unseen data as compared to n-gram language models. Anti-social messages are a significant problem in social media. In this paper we present an approach to classify the natural language text as anti-social text using Random Forest classifier. In this paper we are addressing the challenge to identify anti-social messages using this algorithm using vector ensemble technique to classify anti-social text in offline mode. Word to vector approach has been used for word embeddings to train the model. This paper combines word to vector approach with random forest classifier using a multilayer network.