kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Comparing Prophet, XGBoost, and LSTM Models for Web Traffic Forecasting: Assessing Model Performance Across Various Time Series Forecasting Scenarios
KTH, School of Electrical Engineering and Computer Science (EECS).
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
Jämförelse av Prophet, XGBoost och LSTM modeller för webbtrafikprognoser : En utvärdering av noggrannhet och effektivitet i olika tidsserieprognoskontexter (Swedish)
Abstract [en]

Accurately forecasting website traffic is a crucial task for businesses aiming to optimize resource allocation, reduce costs, enhance user experience, and strategically plan marketing campaigns. This thesis tackles the challenge of selecting the most effective time series forecasting model for predicting web traffic among Prophet, LSTM, and XGBoost, a task crucial for operational efficiency and user satisfaction due to the nonlinear and dynamic nature of web traffic data. Despite its importance, no comprehensive comparison of Prophet, LSTM, and XGBoost models for this purpose exists. To address this, we collected and preprocessed a dataset from Kaggle, implemented and fine-tuned Prophet, LSTM, and XGBoost models, and evaluated them based on RMSE, MAPE, and computational efficiency. Our results demonstrate that the LSTM model outperforms both Prophet and XGBoost in terms of prediction accuracy, particularly for short to medium-term forecasts. The XGBoost model showed improved performance with larger datasets, while the Prophet model excelled in computational efficiency and was ideal for quick iterations. These findings suggest that LSTM is best suited for complex, nonlinear patterns in web traffic data, XGBoost is effective for large-scale datasets, and Prophet is optimal for providing quick forecasts, even if they are slightly less accurate.

Abstract [sv]

Att förutse webbplatstrafik är viktigt för företag som vill optimera resurser, minska kostnader, förbättra användarupplevelsen och planera marknadsföring. Denna examensarbete undersöker vilken tid serie modell, bland Prophet, LSTM och XGBoost, som bäst förutsäger webbtrafik. Trots vikten av detta finns ingen omfattande jämförelse av dessa modeller. Vi samlade in data från Kaggle, implementerade och justerade den för Prophet, LSTM och XGBoost, och utvärderade dem med hjälp av RMSE, MAPE och beräknings effektivitet. Våra resultat visar att LSTM-modellen get bättre resultat jämförense med både Prophet och XGBoost när det gäller prognosnoggrannhet, särskilt för kort- till medellångsiktiga prognoser. XGBoost-modellen visade förbättrad prestanda med större datamängder, medan Prophet-modellen utmärkte sig i beräkningsmässig effektivitet och var idealisk för snabba iterationer. Dessa resultat tyder på att LSTM är bäst lämpad för komplexa, icke-linjära mönster i webbtrafikdata, XGBoost är effektiv för storskaliga datamängder, och Prophet är optimal för snabba, om än mindre exakta, prognoser.

Place, publisher, year, edition, pages
2024. , p. 56
Series
TRITA-EECS-EX ; 2024:482
Keywords [en]
Website Traffic, Time Series Forecasting, Computational Efficiency, Prediction Accuracy
Keywords [sv]
Webb Trafik, Tidsserie Prognoser, Beräkning Effektivitet, Prediktion Noggrannhet
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-351600OAI: oai:DiVA.org:kth-351600DiVA, id: diva2:1887941
External cooperation
Knowit Connectivity
Supervisors
Examiners
Available from: 2024-09-23 Created: 2024-08-09 Last updated: 2024-09-23Bibliographically approved

Open Access in DiVA

fulltext(1456 kB)691 downloads
File information
File name FULLTEXT01.pdfFile size 1456 kBChecksum SHA-512
7991a2f6588b2cc9c32c805faf396b381fdc00ca931f7d257d468db3c7a2a293259c567f108203fce9305d8145e549d27ac79fbc6695eebf5c96d24dfb9a07a6
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 691 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 569 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf