Published Paper Details:

The popularity of Google's video distribution platform, YouTube, has led to an influx of users. Unfortunately, this success has also attracted a significant number of malicious individuals who aim to promote their own videos or distribute viruses and malware. Due to YouTube's limited comment moderation tools, spam comments increase rapidly, often leading video owners to disable the comment section. Detecting and classifying spam comments pose a challenge as they are usually short, filled with slangs, symbols, and abbreviations. In this research paper, we conducted an evaluation of various high-performance classification techniques to detect and analyze spam comments. Through statistical analysis, we found that Decision trees, Logistic regression, Bernoulli Naive Bayes, Random Forests, Linear, Gaussian, and SVM classifiers are all statistically equivalent in terms of their maximum performance rate, with a confidence level of 99.9%. Hence, it is crucial to find a way to identify and report these comments before innocent users encounter them. To address this issue, we propose utilizing Spyder, a popular data analysis tool, to analyze the data and create well-fitted models for the project. These models can then be used to identify patterns and make predictions, which can be integrated into further algorithms to obtain the desired output. Overall, this paper emphasizes the importance of detecting spam comments on YouTube and presents an evaluation of multiple classification techniques as potential solutions.