LE3 - Outlier Analysis

Deletion / Omission

Imputing

rule based

• source: Business / Data understanding

• Gießen is the same as Geißen

• replace(data$spalte, dat$spalte == “Geißen”, “Gießen”) <- ändern

• dat[dat$fixation> min.length, ] <- Filtern

cluster based

• source: unsupervised learning

• build clusters

• look for points far away from cluster

• DBSCAN / KMEANS

regression based

• supvervised learning

  • Regression model

  • look for high residuals

  • special: everything outside of 𝜇 +- 3𝜎

• least square model

usefull to find outliers

usefull to find outliers

unsupervised

usefull to find outliers

supervised