Evaluating the Performance of Machine Learning Classifiers for Detecting Twitter Spam
Dipalee Bhalerao Borse, Swati K. Borse, Vijaya Ahire · International Journal of Computer Applications · 2023
The usage of social networking sites is rising rapidly every day.The popularity of twitter as a microblogging site is huge in normal users as well as illegitimate users.The people with wrong intentions use twitter to spread spam posts which results in phishing, monetary loss, un-useful or noisy data on social media, stealing personal information etc.It becomes extremely important to stop spamming activities.In this paper six machine learning classifiers, which are Logistic regression & Support Vector machine (linear models) and Random forest, K-Nearest Neighbor, Decision tree and Naive Bayes, (nonlinear models), have been implemented on existing data and compared the performance using different parameters such as accuracy, F1-score, recall, precision, f-measure.Among the six classifiers random forest has shown better accuracy followed by K-nearest neighbor classifier for large continuous dataset than small or random dataset.The accuracy is increased from 3% to 13% for large continuous data.Also False positive ratio of random forest and K-nearest neighbor algorithm 0.001 and 0.005 respectively which is much lesser than other algorithms.With lowest accuracy and highest FPR Naive Bayes algorithm performed worst for large datasets.