Chapter 3: Data Preprocessing

Preprocess Steps

Data cleaning

Data integration and transformation

Data reduction

April 19, 2013

Data Mining: Babu Ram Dawadi

Why Data Preprocessing?

Data in the real world is dirty incomplete: lacking attribute values, lacking certain attributes of interest, or containing only aggregate data noisy: containing errors or outliers inconsistent: containing discrepancies in codes or names No quality data, no quality mining results! Quality decisions must be based on quality data Data warehouse needs consistent integration of quality data
Data Mining: Babu Ram Dawadi 2

April 19, 2013

Multi-Dimensional Measure of Data Quality

A well-accepted multidimensional view: Accuracy Completeness Consistency Timeliness Believability Value added Interpretability Accessibility
Data Mining: Babu Ram Dawadi 3

April 19, 2013

Major Tasks in Data Preprocessing

Data cleaning

Fill in missing values, smooth noisy data, identify or remove outliers, and resolve inconsistencies Integration of multiple databases, data cubes, or files Normalization and aggregation Obtains reduced representation in volume but produces the same or similar analytical results

Data integration

Data transformation

Data reduction

April 19, 2013

Data Mining: Babu Ram Dawadi

Forms of data preprocessing

April 19, 2013

Data Mining: Babu Ram Dawadi

Data Cleaning

Data cleaning tasks

Fill in missing values Identify outliers and smooth out noisy data Correct inconsistent data

April 19, 2013

Data Mining: Babu Ram Dawadi

Missing Data

Data is not always available

E.g., many tuples have no recorded value for several attributes, such as customer income in sales data
equipment malfunction inconsistent with other recorded data and thus deleted data not entered due to misunderstanding certain data may not be considered important at the time of entry not register history or changes of the data

Missing data may be due to

Missing data may need to be inferred.

Data Mining: Babu Ram Dawadi 7

April 19, 2013

How to Handle Missing Data?

Ignore the tuple: usually done when class label is missing (assuming

the tasks in classificationnot effective when the percentage of

missing values per attribute varies considerably.

Fill in the missing value manually: tedious + infeasible? Use a global constant to fill in the missing value: e.g., unknown, a new class?! Use the attribute mean to fill in the missing value Use the attribute mean for all samples belonging to the same class to fill in the missing value: smarter Use the most probable value to fill in the missing value: inferencebased such as Bayesian formula or decision tree
Data Mining: Babu Ram Dawadi 8

April 19, 2013

Noisy Data

Noise: random error or variance in a measured variable Incorrect attribute values may due to faulty data collection instruments data entry problems data transmission problems technology limitation inconsistency in naming convention Other data problems which requires data cleaning duplicate records incomplete data inconsistent data
Data Mining: Babu Ram Dawadi 9

April 19, 2013

How to Handle Noisy Data?

Binning method: first sort data and partition into (equi-depth) bins then one can smooth by bin means, smooth by bin median Equal-width (distance) partitioning:

Equal-depth (frequency) partitioning:

It divides the range into N intervals of equal size: uniform grid if A and B are the lowest and highest values of the attribute, the width of intervals will be: W = (B-A)/N. It divides the range into N intervals, each containing approximately same number of samples Managing categorical attributes can be tricky.

Combined computer and human inspection detect suspicious values and check by human
Data Mining: Babu Ram Dawadi

April 19, 2013

10

Cluster Analysis
Clustering: detect and remove outliers

April 19, 2013

Data Mining: Babu Ram Dawadi

11

Regression
y Regression: smooth by fitting the data into regression functions
Y1

Y1

y=x+1

X1

April 19, 2013

Data Mining: Babu Ram Dawadi

12

Data Integration

Data integration: combines data from multiple sources. Schema integration integrate metadata from different sources Entity identification problem: identify real world entities from multiple data sources, e.g., A.cust-id B.cust-# Detecting and resolving data value conflicts for the same real world entity, attribute values from different sources are different possible reasons: different representations, different scales, e.g., metric vs. British units
Data Mining: Babu Ram Dawadi 13

April 19, 2013

Handling Redundant Data in Data Integration

Redundant data occur often when integration of multiple databases

The same attribute may have different names in different databases

One attribute may be a derived attribute in another table.

Redundant data may be able to be detected by correlational analysis

Careful integration of the data from multiple sources may help reduce/avoid redundancies and inconsistencies and improve mining speed and quality
Data Mining: Babu Ram Dawadi 14

April 19, 2013

Data Transformation

Smoothing: remove noise from data Aggregation: summarization, data cube construction Generalization: concept hierarchy climbing Normalization: scaled to fall within a small, specified range

min-max normalization z-score normalization

normalization by decimal scaling

New attributes constructed from the given ones
Data Mining: Babu Ram Dawadi 15

Attribute/feature construction

April 19, 2013

Data Transformation: Normalization

min-max normalization

Min-max normalization performs a linear transformation on the original data.

Suppose that mina and maxa are the minimum and the maximum values for attribute A. Min-max normalization maps a value v of A to v in the range [new-mina, new-maxa] by computing:

v= ( (v-mina) / (maxa mina) ) * (new-maxa newmina)+ new-mina

April 19, 2013

Data Mining: Babu Ram Dawadi

16

Data Transformation: Normalization

Z-score Normalization:

In z-score normalization, attribute A are normalized based on the mean and standard deviation of A. a value v of A is normalized to v by computing:

v = ( ( v

)/ A)

where and A are the mean and the standard deviation respectively of attribute A.

This method of normalization is useful when the actual minimum and maximum of attribute A are unknown.
Data Mining: Babu Ram Dawadi 17

April 19, 2013

Data Transformation: Normalization

Normalization by Decimal Scaling

Normalization by decimal scaling normalizes by moving the decimal point of values of attribute A.

The number of decimal points moved depends on the maximum absolute value of A.
a value v of A is normalized to v by computing: v = ( v / 10j ). Where j is the smallest integer such that Max(|v|)<1.

April 19, 2013

Data Mining: Babu Ram Dawadi

18

Data Reduction Strategies

Warehouse may store terabytes of data: Complex data analysis/mining may take a very long time to run on the complete data set Data reduction Obtains a reduced representation of the data set that is much smaller in volume but yet produces the same (or almost the same) analytical results

Data reduction strategies Data cube aggregation Dimensionality reduction concept hierarchy generation
Data Mining: Babu Ram Dawadi 19

April 19, 2013

Data Compression

String compression Typically lossless Audio/video compression Typically lossy compression, with progressive refinement Sometimes small fragments of signal can be reconstructed without reconstructing the whole

April 19, 2013

Data Mining: Babu Ram Dawadi

20

Data Compression

Original Data
lossless

Compressed Data

Original Data Approximated

April 19, 2013 Data Mining: Babu Ram Dawadi 21

Two Styles of Data Mining

Descriptive data mining

characterize the general properties of the data in the database finds patterns in data and the user determines which ones are important

Predictive data mining

perform inference on the current data to make predictions we know what to predict

Not mutually exclusive

used together Descriptive predictive Eg. Customer segmentation descriptive by clustering Followed by a risk assignment model predictive by ANN
Data Mining: Babu Ram Dawadi 22

April 19, 2013

Descriptive Data Mining (1)

Discovering new patterns inside the data Used during the data exploration steps Typical questions answered by descriptive data mining what is in the data what does it look like are there any unusual patterns what dose the data suggest for customer segmentation users may have no idea which kind of patterns may be interesting

April 19, 2013

Data Mining: Babu Ram Dawadi

23

Descriptive Data Mining (2)

patterns at verious granularities geograph

country - city - region - street university - faculty - department - minor

student

Fuctionalities of descriptive data mining Clustering

Ex: customer segmentation

summarization visualization Association

Ex: market basket analysis

April 19, 2013

Data Mining: Babu Ram Dawadi

24

A model is a black box

X: vector of independent variables or inputs Y =f(X) : an unknown function Y: dependent variables or output a single variable or a vector inputs X1,X2 Model Y output

The user does not care what the model is doing it is a black box
April 19, 2013 Data Mining: Babu Ram Dawadi 25

Predictive Data Mining

Using known examples the model is trained the unknown function is learned from data the more data with known outcomes is available the better the predictive power of the model Used to predict outcomes whose inputs are known but the output values are not realized yet

Never %100 accurate

Its performance on unknown data is much more important
Data Mining: Babu Ram Dawadi 26

April 19, 2013

Architecture: Typical Data Mining System

Graphical User Interface Pattern Evaluation Data Mining Engine Database or Data Warehouse Server
data cleaning, integration, and selection Knowl edgeBase

Database
April 19, 2013

Data World-Wide Other Info Repositories Warehouse Web

Data Mining: Babu Ram Dawadi 36

Architecture: DM system

A good data mining architecture will help to make best use of software environment, perform DM tasks in efficient and timely manner, interoperate and exchange information with other information systems, be adaptable to varying user requirement.
Data mining system architecture includes the consideration of coupling a data mining system with a database or data warehouse system

April 19, 2013

Data Mining: Babu Ram Dawadi

37

Coupling

There are several possible designs such as no coupling, loose coupling, semi tight coupling and tight coupling
No coupling means that a data mining system will not utilize any function of a database or data warehouse system

It may fetch data from a particular source (such as a file system), process data using some data mining algorithms, and then store the mining results in another file
Data Mining: Babu Ram Dawadi 38

April 19, 2013

Loose Coupling

Loose coupling means that a data mining system will use some facilities of a database or data warehouse system Fetching data from a data repository managed by database or data warehouse system, and then storing the mining results either in a file or in a designated place in a database or data warehouse

April 19, 2013

Data Mining: Babu Ram Dawadi

39

Semi tight coupling

besides linking a data mining system to database or data warehouse system, efficient implementations of a few essential data mining primitives can be provided in the database or data warehouse system These primitives can include sorting, indexing, aggregation, histogram analysis, multi-way join, and some pre-computation of some essential statistical measures, such as sum, count max, min, and so on.
Data Mining: Babu Ram Dawadi 40

April 19, 2013

Tight Coupling

Tight coupling means that a data mining systems smoothly integrated into the database or data warehouse system The data mining subsystem is treated as one functional component of an information system This approach is highly desirable since it facilitates efficient implementations of data mining functions, high system performance, and an integrated information processing environment

A well designed data mining system should offer tight or semi tight coupling with a database or data warehouse system.
Data Mining: Babu Ram Dawadi 41

April 19, 2013