On the Detecting Anomalies within the Clickstream Data: Case Study for Financial Data Analysis Websites

Pinar E., Gul M. S., AKTAŞ M. S., Aykurt I.

6th International Conference on Computer Science and Engineering, UBMK 2021, Ankara, Türkiye, 15 - 17 Eylül 2021, ss.314-319 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/ubmk52708.2021.9558965
  • Basıldığı Şehir: Ankara
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.314-319
  • Anahtar Kelimeler: Anomaly detection, Big data, Machine learning, Streaming data
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

In current systems, each user behaves uniquely and within certain behavior boundaries. Behaviors that go beyond these boundaries are called anomaly behavior. Detection of such behaviors is very important, especially in the financial sector. The most important reason for this is that behaviors that are out of normal may also threaten the existence of the system, so they have the potential to cause economic damage to the company. Previous studies have been done without using big data platforms and have problems with big data processing. In our solution, we performed machine learning algorithms on big data platforms for this process. This solution includes extracting features from user behavior data, detecting outliers with the help of boxplot analysis, and then processing these data with machine learning on big data processing platforms. At the end of our study, success and performance comparisons of machine learning algorithms and big data processing platforms were made.