click below
click below
Normal Size Small Size show me how
Exam 1 Part 5
Data Processing and Cleaning
Question | Answer |
---|---|
What is data and attribute? | Data is an object encapsulating a collection of attributes. Attribute is a property/characteristic of data object |
What are the different types of attributes? | Categorial (Nominal, Binary, Ordinal), Numeric (Interval, Ratio, Continuous, Discrete) |
What is the difference between ordinal and nominal attributes? | Ordinal attributes values order objects (<, >) while nominal attributes only distinguish (=, !=) |
What is graph data? Give example. | storage of data in a graph whose structure allows traversal in a particular way to convey additional information about the data. Examples: Generic graph, a molecule, and webpages, binary trees, heaps |
What is noise in data? Give example. | random or irrelevant information that is present in a dataset. Noise can be introduced by a variety of factors, such as measurement errors, outliers, incomplete data, or irrelevant features. For attributes, noise refers to modification of original values. E.g. measurement errors in common measurement equipment or distortion of person's voice on equipment |
What is outlier in data? Give example. | An outlier is an observation that is unusually small or large. It could be an error in recording the value, the point doesn't belong in the sample, or the observation is actually valid. |
Explain the techniques in handling missing values for different types of attributes | Eliminate data object instance, impute values (e.g. mean/mode), or ignore missing values. |
What is the relation between correlation and covariance? Give examples. | covariance measures degree to which 2 different variables vary (i.e. positive means they increase together, negative decrease). correlation is standardized linear relationship between 2 variables (cov/sdxsdy). It ranges from -1 to 1 and shows a standardized strength of the relationship. |
What is curse of dimensionality? | The amount of data needed to describe the space increases exponentially as the number of dimensions grows. |
What are techniques to reduce dimensionality? | PCA, LDA, feature selection, Autoencoder, manifold learning |
Why is it important to reduce dimensionality? | Increasing dimensionality causes the data to become more sparse and occupy more space. becomes difficult to get enough samples to accurately represent the space; leads to overfitting of models and a lack of generalization to new data. Easier to visualized data, reduce noise. |
Explain PCA | identifies the most important features or dimensions in a dataset by projecting the data onto a lower-dimensional space. It does this by identifying the axes that capture the most variance in the data, and then projecting the data onto those axes |
What is binarization? | converting continuous or categorical attributes into one or more binary variables (one-hot encoding) |
What is discretization? | process of converting continuous attributes into ordinal attribute (e.g. pedal width to S,M,L or 0,1,2) |
Why is normalization done to data? | data can have different scales, units, means, ranges. Simplifies data so comparisons are on even footing. |
What is standardization normalization? | Z-score normalization - standardizes attributes to mean of 0 and sd of 1. |
What is min-max normalization? | scales data to fixed range (e.g. 0 to 1). Bad for data w/ outliers |
How to determine if a categorical or numerical attribute is redundant? | If attribute has high collinearity with another - they capture the same feature probably. |
What is entropy-based binning? | Separating data into bins based on the highest information gain |