25.10.2016 Views

SAP HANA Predictive Analysis Library (PAL)

sap_hana_predictive_analysis_library_pal_en

sap_hana_predictive_analysis_library_pal_en

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

3.2.14 Support Vector Machine<br />

Support Vector Machines (SVMs) refer to a family of supervised learning models using the concept of support<br />

vector. Compared with many other supervised learning models, SVMs have the advantages in that the models<br />

produced by SVMs can be either linear or non-linear, where the latter is realized by a technique called Kernel<br />

Trick.<br />

Like most supervised models, there are training phase and testing phase for SVMs. In the training phase, a<br />

function f(x):->y where f(∙) is a function (can be non-linear) mapping a sample onto a TARGET, is learnt.<br />

The training set consists of pairs denoted by {x i , y i }, where x denotes a sample represented by several<br />

attributes, and y denotes a TARGET (supervised information). In the testing phase, the learnt f(∙) is further<br />

used to map a sample with unknown TARGET onto its predicted TARGET.<br />

In the current implementation in <strong>PAL</strong>, SVMs can be used for the following three tasks:<br />

●<br />

●<br />

●<br />

Support Vector Classification (SVC)<br />

Classification is one of the most frequent tasks in many fields including machine learning, data mining,<br />

computer vision, and business data analysis. Compared with linear classifiers like logistic regression, SVC<br />

is able to produce non-linear decision boundary, which leads to better accuracy on some real world<br />

dataset. In classification scenario, f(∙) refers to decision function, and a TARGET refers to a "label"<br />

represented by a real number.<br />

Support Vector Regression (SVR)<br />

SVR is another method for regression analysis. Compared with classical linear regression methods like<br />

least square regression, the regression function in SVR can be non-linear. In regression scenario, f(∙)<br />

refers to regression function, and TARGET refers to "response" represented by a real number.<br />

Support Vector Ranking<br />

This implements a pairwise "learning to rank" algorithm which learns a ranking function from several sets<br />

(distinguished by Query ID) of ranked samples. In the scenario of ranking, f(∙) refers to ranking function,<br />

and TARGET refers to score, according to which the final ranking is made. For pairwise ranking, f(∙) is<br />

learnt so that the pairwise relationship expressing the rank of the samples within each set is considered.<br />

Because non-linearity is realized by Kernel Trick, besides the datasets, the kernel type and parameters should<br />

be specified as well.<br />

Prerequisite<br />

No missing or null data in the inputs.<br />

SVMTRAIN<br />

This function reads the input data and generates training model.<br />

Procedure Generation<br />

CALL SYS.AFLLANG_WRAPPER_PROCEDURE_CREATE (‘AFL<strong>PAL</strong>’, ‘SVMTRAIN’,<br />

‘’, '', );<br />

<strong>SAP</strong> <strong>HANA</strong> <strong>Predictive</strong> <strong>Analysis</strong> <strong>Library</strong> (<strong>PAL</strong>)<br />

<strong>PAL</strong> Functions P U B L I C 227

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!