CYBER OFFENSIVE PREDICTIVE ANALYSIS ON TWITTER DATA USING MACHINE LEARNING

CYBER OFFENSIVE PREDICTIVE ANALYSIS ON TWITTER DATA USING MACHINE LEARNING

Sanjay R 1 , Gautam R 2 , Mrs.Jayashankari J 3 , Dr.Preetha M 4

Student, Information Technology, Prince Shri Venkateshwara Engineering

College 1,2

Assistant Professor, Department of Information Technology 3

Professor, Department of Computer Science and Engineering 4

Prince Shri Venakteshwara Padmavathy Engineering College

Abstract:

Cyber-bullying refers to the utilization of aggressive, violent, or offensive language, targeting a selected cluster of individuals sharing a typical property, whether or not this property is their gender, their grouping or race, or their beliefs and faith. With the rising of social networks, communication between folks from completely different cultural and psychological backgrounds has become a lot direct, leading to a lot of ‘‘cyber’’ conflicts between these folks and it's become a significant downside. Therefore, arises the need to discover such speech mechanically in the Twitter victimization Machine Learning rule. The system depends on the detection of 3 basic tongue elements equivalent to Insults, Swears, and person. Here is the classification of, the real-world dataset from Twitter, one in every of the highest 5 networks with the highest share of cyber-bullying instances. An arrangement named Random Forest, call tree and logistical regression has been utilized to discover the incidence of cyber entities in Twitter. These algorithms are an effective method of Decision Making because they: Clearly lay out the problem so that all options can be challenged, allow us to analyze fully the possible consequences of a decision, and provide a framework to quantify the values of outcomes and the probabilities of achieving them. The call tree classifier can be used in both classification and regression. It can help represent the decision as well as make a decision. The Random Forest classifier is consisting of multiple decision tree classifiers. Each tree gives a class prediction individually. The maximum number of the predicted class is our final result. The projected system prediction model is to use a text classification approach that involves the development of machine learning classifiers from labeled text instances. Datasets containing bullying texts, messages or posts are collected and prepared for the machine learning algorithms using natural language processing. In our projected model, at first information preprocessing is completed and so tokenization and normalization can happen. The processed datasets are then used to train the machine learning algorithms for detecting any harassing or bullying message on social media including Facebook and Twitter. The potency and accuracy are high within the projected model due to multi-model algorithms specifically random forest, call tree and logistical regression. Additionally, this model may acknowledge all sorts of matter inputs and predict the output.

Keywords: Regression  algorithm, Naïve Bayes Algorithm, Decision tree Algorithm, Django, Flask