Professional Documents
Culture Documents
Metode Pohon Regresi Untuk Eksploratori Data Dengan Peubah Yang Banyak Dan Kompleks
Metode Pohon Regresi Untuk Eksploratori Data Dengan Peubah Yang Banyak Dan Kompleks
ABSTRACT
Regression trees are used to predict membership of cases or
objects in the classes of a categorical dependent variable from
their measurements on one or more predictor variables.
Regression tree analysis is one of the main techniques used in
so-called data mining. The goal of regression trees is to predict
or explain responses on a categorical dependent variable. The
flexibility of regression trees make them a very attractive
analysis option, but this is not to say that their use is
recommended to the exclusion of more traditional methods.
Indeed, when the typically more stringent theoretical and
distributional assumptions of more traditional methods are met,
the traditional methods may be preferable. But as an exploratory
technique, or as a technique of last resort when traditional
methods fail, regression trees are, in the opinion of many
researchers, unsurpassed. This research used data from survey
on farmer income conducted by BPS-Statistics Indonesia (for
Jawa Timur Province) in 2004, and regression method based on
tree structure with CART algorithm to build a model. The results
show that farmer’s income is interconnected with expenditure of
farming activities and land ownership. Despitefully, there are
other non-technical factors that also can influence the income.
This factors among others the social condition of pertinent
agriculture household, for example, education level, age and
also other external factors such as soft loan from government
and agriculture counseling. These matters indicate that the
earnings from farming activities is represented by the function of
those factors.
METODOLOGI
CART (Classification and Regression Trees) adalah salah
satu metode atau algoritma dari salah satu teknik eksplorasi
data yaitu teknik pohon keputusan. Metode ini dikembangkan
oleh Leo Breiman, Jerome H. Friedman, Richard A. Olshen dan
Charles J. Stone sekitar tahun 1980-an. Menurut Breiman et al.
(1993), CART merupakan metodologi statistik nonparametrik
yang dikembangkan untuk topik analisis klasifikasi, baik untuk
peubah respon kategorik maupun kontinu. CART menghasilkan
suatu pohon klasifikasi jika peubah responnya kategorik, dan
menghasilkan pohon regresi jika peubah responnya kontinu.
Tujuan utama CART adalah untuk mendapatkan suatu kelompok
data yang akurat sebagai penciri dari suatu pengklasifikasian.
Bentuk dari CHART adalah seperti berikut ini :
node/simpul
A
Ya tidak
cabang
x1≤ α ?
C
B
Ya tidak
x2 ≤ β ?
C C
Simpul akhir
[
JKS (t ) = ∑ ( yi ( t ) − y(t ) ]
2
dengan i = 1,2,…, N t
xn ∈t
DAFTAR PUSTAKA
BPS. 2004. Pedoman Teknis BPS Propinsi dan BPS
Kabupaten/Kota. Sensus Pertanian 2003. BPS, Jakarta.
BPS. 2004. Survei Pendapatan Petani: Pendapatan Rumah
Tangga Pertanian. Sensus Pertanian 2003. BPS, Jakarta.
Breiman L, Friedman J.H., Olshen R.A., and Stone C.J. 1993.
Classification and Regression Trees. Chapman and Hall.
New York.
Soekartawi. 2002. Prinsip Dasar Ekonomi Pertanian : Teori dan
Aplikasi. PT. RajaGrafindo Persada, Jakarta.
Statsoft. 2003. Classification and Regression Trees (C&RT).
[terhubung-berkala]
http://www.statsoft.com/textbook/stcart.html
[10 Maret 2005].