Data Preprocessing

Before measuring similarity, each data item is first converted into a:

What is the Euclidean distance between the points (1, 2) and (4, 5)?

For categorical features with m matches out of p features, the proximity distance is:

Which similarity measure is preferred for high-dimensional, sparse, unstructured data?

A cosine similarity value close to 1 means the two items are:

Feature discretization converts:

To impute a missing CATEGORICAL attribute value, you typically use the:

Why is normalization important for distance-based algorithms?

Which normalization is preferred when min/max are unknown or outliers dominate?

A collection of related points that is anomalous, though its individual points are not, is a:

Similarity between two data items is a real-valued score.

Euclidean distance is the straight-line distance between two points.

Cosine similarity is strongly affected by document length (vector magnitude).

Unstructured-data feature vectors are usually sparse with many zeros.

Discretization maps each interval of a continuous feature to a categorical value.

The median is good against noise when imputing numeric missing values.

An attribute with a larger range can drift/dominate the similarity measure.

Min-max normalization rescales values into a fixed range such as [0,1].

A contextual outlier is anomalous only within a certain context.

In distance-based detection, outliers have a dense neighborhood.

Match each numeric distance/measure.

Match each feature type with its similarity approach.

Match each preprocessing step.

Match each normalization method.

Match each outlier concept.