Java Data Mining Concepts - Java Data Mining: Strategy, Standard, and Practice - page 153

Table 7-7 Real-estate appraisal data

Attribute
type

Usage
type Attribute name Logical name Preparation

LOAN_ID Loan Id Inactive

CITY City Categorical Active Prepared

COUNTY County Categorical Active Prepared

STATE State Categorical Active Prepared

HOME_SIZE Size of the home in sq. ft Numerical Active Not prepared

YEAR_BUILT Year home was built Numerical Active Not prepared

LAND_SIZE Total land size in sq. ft Numerical Active Not prepared

NUM_ROOMS Number of rooms Numerical Active Not prepared

NUM_GARAGES Number of garages Numerical Active Not prepared

POOL_TYPE Type of pool (none,

inground, above ground)

Categorical Active Prepared

SCHOOLS Number of schools Categorical Active Prepared

NUM_MALLS Number of nearby malls Numerical Active Not prepared

CRIME RATE City crime rate Numerical Active Not prepared

SCHOOL RATING Avg. school ratings Numerical Active Not prepared

APPRAISAL_VALUE Home appraisal value Numerical Ta rg e t

such as city, county, state, house size, year built, land size, number of
rooms, garages, pool type, number of schools, school rating, number
of nearby malls, crime rate, and the target appraisal value.

7.2.4 Select Algorithm: Find the Best Fit Algorithm

The algorithms discussed for classification in Section 7.2.1 can also be
used for regression. However, among them, support vector machine
(SVM), feed forward neural networks, and decision tree algorithms
are commonly used. There are other regression algorithms, such as
linear regression and generalized linear models (GLM) that are not cur-
rently included among the JDM supervised algorithms. Regression
model quality depends on data characteristics, algorithm choice, and
settings to find the most appropriate algorithm, or rely on selection
performed by the data mining engine.