predictive analytics applying it in your organization - bocx 2014

27
© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary © 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary Predictive Analytics Applying It In Your Organization

Upload: martin-sykora

Post on 13-Apr-2017

223 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Predictive AnalyticsApplying It In Your Organization

Page 2: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Agenda• Introduction to Predictive Analytics• SAP Predictive Analysis Overview• Demo Scenarios• Q&A• Getting Help• Appendix

Page 3: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Introduction to Predictive Analysis

Page 4: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Why Use Predictive AnalyticsPredictive analytics encompasses a variety of statistical techniques from modeling, machine learning, and data mining that analyze current and historical facts to make predictions about future, or otherwise unknown, events.

Page 5: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Extend Analytics Capabilities

Page 6: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

SAP Predictive Analysis Overview

Page 7: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

SAP Predictive Analysis

Acquire data from various data sources, • Excel• CSV• SAP

HANA• Universe• RDBMS

Perform data enrichment operations,• Data

manipulation• Data

cleansing• Semantic

enrichment

Get insight into data prediction using statistical algorithms,• Clustering• Decision

Tree• Trending• Associatio

n• Regression

Build visualization using wide selection of charts and graphs.

Explore and analyze row and predictive datasets using visualizations.

Publish and share predictive datasets to,• SAP HANA• SAP

Explorer• SAP

Streamwork

• Excel• CSV• RDBMS• Email• PDF

Page 8: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Predictive Analysis Architecture• The R open source statistics language and its

3,500 plus packages or algorithms is the most popular predictive analysis tool according to the Rexer Analytics 2011 Data Mining Survey, close to half of all data miners (47%) using it.

• R resides on a separate server, side by side with SAP HANA. Data stored in SAP HANA tables is passed to the R server and transformed into R data vectors or data frames, which is the data format used by R. R script code is embedded in SAP HANA SQLScript and is also passed over to R for processing on the R server. The results, in the form of R data vectors or data frames, are transferred back to SAP HANA, which then converts it back to an SAP HANA table. All the communication, data transfer and transformation tasks are performed by the SAP HANA Platform.

Page 9: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Predictive Analysis Algorithms

Supervised Learning: Target information and examples (training sets) provided to the learner (component)

• Unsupervised, no training sets• Find patterns in transactional datasets for generating

association rules• Predict future transacting behavior

Association• Unsupervised, no training sets• Cluster observations into groups without prior knowledge of

those relationships• Input to decision making processCluster• Supervised, training sets for error correction• Predict class membership through probability• Scoring models for churn, defect, triageClassification

• Structured, training sets, k+1 forecast• Predicts the value of a variable based on supplied data• Forecasting with or without trending or seasonal adjustmentsRegression• Unsupervised, no training sets• Find objects that do not comply with the general behavior or

model• Churn, fraud, fault analysisOutliers

Page 10: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Create New Custom-R Component

Page 11: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Demo Scenarios

Page 12: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Customer Segment Analysis - RetailBusiness Scenario• A consumer clothing retail company wants to

perform a targeted marketing campaign to increase particular product sales

• Marketing team wants to understand the demographic of their customer

• Company wants to target specific customer segment to achieve maximum success from the campaign

Page 13: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Targeted Customer Offering - InsuranceBusiness Scenario• An insurance company wants to increase their

revenue by extending product offering to targeted customers

• Company wants to know the likelihood of their customers to buy particular product offering

• Company wants to focus their agents sales initiatives to targeted customers

Page 14: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Hospital Staffing Forecast - HealthcareBusiness Scenario• Hospital management team wants to manage

staffing assignments based on forecasted patient demand

• Historical statistics for Available Beds, Occupied Beds, Inpatient Admissions and Day Case Admissions by month is available

• Management wants to use the historical data to predict and forecast hospital staffing needs for next 5 years

Page 15: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Customer Retention - BankingBusiness Scenario• Identify the attributes of those customers who have

been retained, in contrast to those who have left or churned.

• Develop strategies to maximize the retention of customers.

Page 16: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Q & A

Page 17: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Getting Help• SAP Analytics Training on youTube

o Create an Analysis: SAP Predictive Analysis 1.14o Create a Custom R Component: SAP Predictive Analysis 1.14o Install and Configure Open Source R Algorithms: SAP Predictive Analysis 1.14o Save, Export and Import Data Models: SAP Predictive Analysis 1.14

SAP BigData http://www.sapbigdata.com/analytics/predictive/SAP BI Blog http://sapbiblog.com/category/predictive-analytics/SAP Community Network

http://scn.sap.com/community/predictive-analysis

SAP Help Portal https://help.sap.com/paSAP HANA http://www.saphana.com/community/learn/solutions/pr

edictive-analysisSAP Insider http://sapinsider.wispubs.com/Search?query=SAP%20Predictive%20AnalyticsSAP Support Portal http://service.sap.com/xsearch

SAP Training https://www.youtube.com/user/SAPAnalyticsTraining

Page 18: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Appendix

Page 19: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA 1.15 What’s New

Page 20: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA 1.15 PAM – Desktop & Server

Page 21: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA 1.15 PAM – Database Connectivity

Page 22: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA Supported AlgorithmsType Brief Algorithm R

Association Find associations between entities based on transactions involving them (aka Market Basket Analysis)

Apriori R-Apriori

Cluster Group individuals in a population together by their similarity, but not driven by any specific purpose

K-MeansSelf-Organizing Maps

R-K-Means

Classification Predict by individual element, which mutually exclusive class this individual belongs to

ABC AnalysisKNNNaïve BayesWeighted Score Analysis

-

Decision Trees

Classification of individual elements through a hierarchical form – useful for model scoring

C4.5CHAID

R-CNR Tree

Regression Classification to estimate or predict, for each individual, the numerical value of some variable

Exponential Geometric Logarithmic Logistic Multiple Linear Polynomial

R-Multiple Linear Regression

Time Series A time series is set of consecutive points collected in uniform time intervals. Smoothing provides forecasting abilities based on collected averages, double with trend, triple with seasonality.

Exponential Smoothing:DoubleSingleTriple

R-Triple Exponential Smoothing

Outlier Material difference from or inconsistent with the remaining set of data, are called anomalies or outlier

Anomaly DetectionInter-quartile RangeVariance Test

-

Page 23: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA Algorithm DescriptionsTime Series Analysis• In this group we use past data points as the basis for

projecting future ones. The data is a variable, such as sales or headcount or telemetry readings, with a series of values over time. We use historical patterns in the known data to make predictions about future values.

Association Analysis• This class of applications looks for associations

between things, with the best known example being the analysis of shopping baskets and the prediction that people who purchased this product are most likely to purchase these other products. It is the analysis is of a basket of items. Other examples include the analysis of telecom services purchased by subscribers; financial policies held by households; telephone calling patterns; visitor paths through a website.

Page 24: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA Algorithm Descriptions (Cont.)Cluster Analysis• This class of application is concerned with grouping the

data into clusters or segments that have similar attributes. It is often used to subset a large data set, as we may be able to better understand the attributes of the smaller subsets. We also cluster data so that we can focus on specific groups within a dataset. The best example of this is customer segmentation where we want to develop specific sales strategies for particular customer groups.Classification Analysis

• This is the largest group of applications of predictive analysis. To predict a variable using the data of other variables that we believe affect the values of the variable that we are trying to predict. The variable that we wish to predict is variously known as the output variable, the target variable, or the dependent variable. The latter name comes from the assertion that it depends on the values of the independent variables or input variables on Analysis.

Page 25: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA Algorithm Descriptions (Cont.)Outlier Analysis• This class of application of predictive analysis looks for

the unusual and the unexpected: unusual in the sense that values in a data set are very different from other values, and unexpected in the sense that the values are very different from what was predicted. The outliers could be as a result of errors in the data or genuine variations in the data, which the predictive model needs to take into account. Outlier or anomaly detection has a major application in the area of fraud detection.

5. Outlier Analysis

Page 26: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

PA Task to Algorithm Sample

Page 27: Predictive analytics   applying it in your organization - bocx 2014

© 2014 Broadstreet Data & NexJ Systems Inc. Confidential and Proprietary.

Contact UsWebsite: Email:

Connect with Us

Thank You