Published Paper Details:

In the days of the Internet, millions of users are using billions of websites in every fraction of seconds. The performance of each and every website depends upon some crucial factors. These factors are called web-metrics. It is crucial to evaluate them for better performance, security etc. These web-metrics consist of web-page load time, number of redirections, total number of pages, First Contentful Paint(FCP), Largest Contentful paint(LCP), Speed Index, DOM size, QLS etc. By evaluating these features we can understand how a website will perform during user usage. Also by evaluating them we can make necessary changes to improve the performance of the website. A huge amount of study is already going on by various researchers. In our paper we made a comparative study by using machine learning models like linear regression, SVM, Random Forest. Also we proposed a model by imposing an ensemble learning method with ML models which can predict the website's load time. Our model can also be able to give information about other web-metrics by evaluating this predicted web-page load time and the corresponding website URL. Our proposed model is directly trained by extracting web-metrics from the input URL. This hybrid model can be used by developers, Administrators to optimize and improve the performance of the corresponding website. Our proposed model is achieving a highly accurate(98%) predicted load time for websites and can successfully predict the other parameters of web-metrics. Also the proposed model is highly flexible to add or remove other web-metrics.