DS - Preprocessing

Why do we need to preprocess data? (2)

• To reduce data size
• To eliminate missing, noisy, or conflicting data
• To improve data quality
• To increase data complexity

What types of data need preprocessing? (3)

• Numeric data
• Categorical data
• Distributed data
• Text, images, audio, video data

What are the consequences of poor-quality data? (3)

• Wrong decisions
• Misleading analysis results
• Inability to integrate data
• Data becomes more complex

How can missing data be handled? (3)

• Delete missing data
• Fill with the mean value
• Fill with most likely value
• Replace with a constant

What is the purpose of binning in noise reduction? (3)

• Sort and divide data into equal-width bins
• Divide data into equal-depth bins
• Reduce noise by mean, median, margin
• Increase data size

Which method is inefficient for handling missing data? (1)

• Delete missing entries
• Fill missing values manually
• Automatically fill using statistics
• Use prediction models

Which methods can be used for noise reduction? (3)

• Binning
• Clustering
• Regression
• Data integration

How is noise handled using regression? (2)

• Divide data into groups
• Fit data to a regression function
• Remove outliers
• Delete inconsistent data

Which methods handle data conflicts? (2)

• Integrating data from multiple sources
• Identifying and removing duplicates
• Normalizing and mapping data using metadata
• Removing unnecessary attributes

What are the steps in data integration? (3)

• Selecting necessary data
• Matching data schema
• Removing duplicates
• Creating new attributes