A Systematic and Comprehensive Study on Machine Learning and Deep Learning Models in Web Traffic Prediction

Abstract

The practice of predicting the traffic that is headed toward a specific website is known as web traffic prediction. To govern a network, network traffic forecasting is crucial. Since clients could experience long wait times and leave a website without a suitable demand prediction, web service providers must evaluate web traffic on a web server very carefully. It is an objective that predicting network traffic is a proactive way to assure safe, dependable, and high-quality network communication. The aim of this paper is to find out the algorithms that can be best fitted for web traffic prediction. If the traffic is more than the server can handle, then it will show error to the people who are reaching the website. So, it becomes difficult to handle a large amount of traffic. One option is we can increase the number of servers but for this to know how many servers should be increased we have to forecast the web traffic. This is one of the applications of web traffic forecasting. To improve traffic control decisions, it is necessary to estimate future web traffic. In this paper, we have discussed the most efficient algorithms that can be utilized for web traffic prediction. Here, SVM, LSTM, and ARIMA are discussed which are comparatively more efficient and optimized algorithms. Many algorithms can be used to predict this website traffic, but the algorithms discussed in this paper are found to be more optimized. So, overall this algorithm can be used for website prediction with great efficiency. These algorithms are found to be quite fast as compared to others and they also give a good accuracy score. So, the results show that the prediction precision is high if these algorithms are utilized.