Predictive Analysis in SAP HANA

Predictive Analysis in SAP Hana

SAP HANA offers a radical new approach to predictive analytics whereby the volume of data that may be analyzed and the speed of the analysis bring a new perspective to predicate analysis.

In database predictive analysis coupled with a modern user interface to define the analysis process and support – the embedding of the analysis process in applications is a powerful combination.

Overview of predictive analysis

Data mining is the exploration and analysis by automatic or semi-automatic means, of layer quantities of data in order to discover meaningful patterns and rules.

The process of data access, data exploration, data prep action, and modelling model deployment.

In-Memory data mining & statistical analysis

In their ways to implement in-memory data mining & statistical analysis

SQL Scripts

The SQL script is an asset of SQL extrusions which allow developers to push data-intensive logic into the database in order to avoid massive data copies to the application server and to leverage sophisticated parallel execution strategies of the database

The set of SQL extensions for the SAP HANA databases which allow developers to push data-intensive logic into the database is called SQL script

These extensions are keys to avoiding massive data copies to the application server and leveraging sophisticated parallel execution strategies of the database.

SQL script V2 supports stored procedures, which provides enhanced control flow capabilities and is positioned to be more scriptable for pushing complex parts of application logic to the database.

It can meet some requirements for reporting like join, aggregation, Etc. when it comes to data minify and statistic analysis, SQL script & is not suitable for implementing complex algorithms

R

Through the ‘R’ integration solution, developers can leverage open Source RS 3000+ External packages to perform wide–range data mining and statistic analysis

An open-source software language and environment for statistical computing and graphics with 3000 add–on packages

The packages cover a wide range of topics

• Cluster analysis 4 final mixture models
• Probability distributions
• Computational econometrics
• Empirical finance
• Statistical generics
• Graphic displays, dynamic graphics graphic devices & visualization
• Machine learning & statistical learning
• Medical image analysis
• Multi variant statistics
• Natural languages processing
• Statistics for the Social – sciences
• Time services analysis

BFL (Business Function Library)

BFL is the calculation library for the applications built on top of the SAP HANA database. The business functions are written in C++ and executed in a database calculation engine. BFL has a road map for data mining and statistical algorithms.

Usage scenario: 1

Utilize open-source ‘R’ as the data mining calculation engine

The application developer can embed R scripts in the SQL script.

Sample codes in SAP HANA SQL Script

Create function LR (in input SVCC – PREC – type, OUT OUTPUTO R_COEF _ type)

LANGUAGE R LANG AS ‘’

CHANGE _FREQ <-input $ CHANGE _ FREQ:

SUCC _ PREC <- input $ SUCC – PREC;

Coefs ,- coef (glm (succ _PREC ~ CHANGE – FREQ, family = position));

CHANGE FREQ <- CICFS [ “(CHANGE _ freq”];

RESCSST <- as.data.frace (cbind (INTERCEPT, CHANGE FREQ))

‘’’;

TRUNCATE TABLE r-coef_tab;

CALL LR (SUCC – PREC – tab r-coef – tab);

SELECT* FROM v-coef – tab;

Scenario 2

Use open source R’s console to interact with SAP HANA data

The diagram is as will as scenario 1:

File edit view terminal help

Lom 787572e:/use/sap/HDB/HDB06>R.Q

>Library (ron)

>Library (RJDBC)

Loading required package : DBI

Loading required package: rjawa

>Jdc path ,-“use/sap/HDB/HDBDD/exc/imprsjdbc.jas”

>drv <-JDBC (“com.ptime.sql.Driver”, jdbe path, identifie-identifier.quote= “ ‘ “)

>conn <-db connect (drv, “jdbc : ptime : local host : 300’s;

“system”  , “manager”)

>SQL <- “select * from TEST – DATA – INT”

>get data frome (com, SQL, “VAr1”)

To register  ‘var1’ with slim ID 3391796252027132

>dim (var 1)

[1[ 10000                    100

>var 1$ col 1[1:10]

[1] 1451      1325     1959     1743     885        1484                     1293

>resuits       <-kmoans (vat2.5)

>results $ size

[1] 2028 2000 2019 1981 1972

>cleanup object (“var1”)

[[1]]

[1] “var”