Flashcards
Data
Cards in this deck (15)
- 1Scalar attributesNominal, Ordinal, Interval, Ratio
- 2NominalProvides only enough information to distinguish one object from another
(==, =/=) ex) zip codes, employee ID, eye color - 3OrdinalValues provide enough information to order objects (<,>) ex) grades, {good, better, best}, 1-5 etc
- 4IntervalThe differences between values are meaningful, i.e. unit of measurement exists (+, -) ex) calendar dates, Celsius etc
- 5RatioBoth differences and ratios are meaningful (/, *) ex) Kelvin, age, electrical current
- 6Discrete attributesHas only a finite or countable set of values, often represented as integer values
ex) zip codes, counts, set of words in a document - 7Continuous attributesReal numbers as attributes typically represented as floats
ex) temperature, height, weight - 8Data quality problemsNoise, outliers, duplicates, missing values
- 9AggregationCombining two or more objects (or attributes) into a single object (or attribute)
Data reduction -> less variability = more stable - 10SamplingEntire set of data is too expensive and/or time consuming to work with
Representative sample = same properties as original set
Larger sample size = more accurate but loses point of sampling - 111. Simple random 2. Without replacement 3. With replacement 4. Stratified sampling1. Equal probability for all
2. Selected item removed from population
3. Selected item not removed from pop.
4. Split data into partitions, then draw randomly from each partitions - 12Curse of dimensionalityWhen dimensionality increases, data becomes more sparse, so density and distance, which is important for some methods, means less
- 13Feature subset selectionRemove redundant and irrelevant features
Done through: brute force, embedded, filter, wrapper - 14Feature creationCreate new attributes that capture important information more efficiently than original data set
- 15Attribute transformationMaps entire set of values of given attribute to a new set such that old value can be identified with new value
Information
- Description
- Data mining Flashcards on Data, created by Johan Isaksson on 23/10/2018.
- Created by
- Johan Isaksson
- Views
- 1
- Rating