Analysis of Website Traffic Time Series Forecasting using ARIMA, Prophet, and LSTM RNN - Abstract and Method Comparison

Sep 13, 2026

Upload by Aran

Rapid internet growth has expanded website traffic data and increased the need for reliable forecasting to support planning and decision-making. This study evaluates multiple approaches for website traffic time series prediction, including ARIMA, Prophet, LSTM, and a hybrid LSTM-GRU RNN model, using the Wikipedia Pageviews Dataset via API and Google Analytics data. Experiments assess how well each method captures trends and seasonality, reporting that LSTM-GRU with 50 epochs achieves the lowest MSE and RMSE values.

LanguageEnglish

CategoryPresentations

FormatPDF

Pages11

Reading Time4 min

AccessFree to download

INTRODUCTION

RELATED WORK

Forecasting Website Traffic Using Prophet Time Series Model

Web Traffic Time Series Forecasting Using LSTM Neural Networks with Distributed Asynchronous Training

Time-Series Analysis with Smoothed Convolutional Neural Network

Analysis of Website Traffic Time Series Forecasting using ARIMA, Prophet, and LSTM RNN Suraj Katwal 1 , Rubim Shrestha2 , Dr. Gajendra Sharma3 1MCA, Kantipur City College, Dept. of Science and Technology, Kathmandu 44600, Nepal 2Assistant Professor, Kantipur City College, Dept. of Science and Technology, Kathmandu 44600, Nepal 3Associate Professor, Kathmandu University, Dept. of Computer Science and Engineering School of Engineering, Dhulikhel 45200, Kavre, Nepal Abstract The rapid growth of the Internet has led to a vast increase in website traffic data. Accurately forecasting website traffic is essential for informed decision-making and future planning. This study comprehensively analyzed four methods for forecasting website traffic time series data: Autoregressive Integrated Moving Average (Hereafter ARIMA), Prophet, Long Short-Term Memory (Hereafter LSTM), and Hybrid Long Short-Term Memory -Gated Recurrent Unit Recurrent Neural Network (Hereafter LSTM-GRU RNN) . The study used Wikipedia Pageviews Dataset using API and Google Analytics data to train and test the forecasting models. The empirical analysis evaluated the accuracy and ability of each method to capture trends and seasonality. The results showed that the LSTM-GRU model with 50 epochs has the lowest MSE value of 0.0057022 and the lowest RMSE value of 0.075513. The LSTM model with 100 epochs has a low MSE value of 0.0057916 and a low RMSE value of 0.0761028, comparable to the LSTM-GRU model with 50 epochs. Keywords: Web Traffic; ARIMA; Prophet; LSTM RNN; LSTM-GRU; Prediction; Time Series 1. INTRODUCTION The increasing population has led to a greater familiarity with the internet worldwide. This has resulted in increased traffic almost everywhere. As a consequence, the company is experiencing a number of problems due to the increased traffic to its website. [4] . The company should invest additional resources to accommodate the increased traffic. This plan could include investing in more servers. The traffic report analyzed earlier can be used to determine if a website management technique plan needs to be implemented to reduce the risk of such mishaps [2,4] . One way to reduce or handle such problems in the future is by forecasting web traffic. Web traffic forecasting is the process of estimating the future size of a web traffic segment or group. This can be done by analyzing historical data, tracking online behavior, or using modeling techniques [7] . Web traffic time series forecasting uses historical web traffic data to predict future web traffic patterns [8] . This can help businesses decide about website design, capacity planning, and marketing campaigns. Web traffic forecasting can be used for various purposes, including estimating website traffic, understanding website trends, and forecasting future traffic volumes [9] . Several methods can be used to forecast web traffic, including time series and regression analysis. In order to determine the best forecast method, we must consider the type of data and the degree of accuracy we want [10] . Time-series analysis is a well-established approach for forecasting web traffic, as it can capture short-term patterns and long-term trends [11] . Regression analysis can identify relationships between different factors and web traffic [12] . In addition to creating more complex models, machine learning can also adapt to changing data over time. 2. RELATED WORK 2.1 Forecasting Website Traffic Using Prophet Time Series Model This paper proposes a method for forecasting website traffic using the Prophet time series model. The prophet time series model is a tool that can be used to predict future events based on past data. The authors argue that the prophet model is well suited for forecasting website traffic because it can handle seasonality and trend changes. The authors collected data on website traffic from Google Analytics and it was used to train the prophet model. The model was then used to predi

FAQs

What dataset and data sources are used to train and test the forecasting models?

The study uses the Wikipedia Pageviews Dataset via API and Google Analytics data to train and test the forecasting models.

Which forecasting methods are compared in this study?

The paper compares ARIMA, Prophet, LSTM, and a hybrid LSTM-GRU RNN approach for website traffic time series forecasting.

How is model performance evaluated?

Performance is evaluated by measuring how accurately each method captures trends and seasonality, using error metrics such as MSE and RMSE.

Sep 13, 2026

Upload by Aran