Release Notes Version 7.0A

©

2013 Pitney Bowes Software Inc. All rights reserved.

This document may contain confidential and proprietary information belonging to Pitney Bowes Inc. and/or its subsidiaries and associated companies. Portrait Software, the Portrait Software logo, Portrait, and Portrait Software's Portrait brand are the trademarks of Portrait Software International Limited and may not be used or exploited in any way without the prior express written authorization of Portrait Software International Limited. Acknowledgement of trademarks Other product names, company names, marks, logos and symbols referenced herein may be the trademarks or registered trademarks of their registered owners. Portrait Software Support If you need help with something that is not covered by this documentation, try the Knowledge Base on our web site. http://support.portraitsoftware.com and follow the links to your product. You can also download other Portrait Software documentation from the site. If you do not have a username and password—or you have forgotten them—please contact us through one of the channels below. If you find a problem with the use, installation, or documentation of this product, please contact us using any of the following methods: Email:[email protected] Phone • USA/Canada 1-800-335-3860 (toll-free) • Rest of world +44 800 840 0001 When you report a problem, it helps if you can tell us: • • • •

The name of the software application The circumstances in which the problem arose What error messages you saw (if any); The version of the software that you were using.

Pitney Bowes Software Inc. June 13, 2013

Contents Chapter 1: Introduction..............................................................................................5 Purpose................................................................................................................................6 Intended audience...............................................................................................................6 Related documentation.......................................................................................................6

Chapter 2: Release overview.....................................................................................9 Product overview...............................................................................................................10 Highlights...........................................................................................................................10

Chapter 3: Release details.......................................................................................11 Scorecard builder changes..............................................................................................12 Support for Unicode..........................................................................................................12 Regular expression changes............................................................................................13 Installation changes..........................................................................................................13

Chapter 4: Installation considerations....................................................................15 Upgrading to Portrait Miner 7.0A.....................................................................................16 Supported platforms.........................................................................................................16 Supported languages........................................................................................................18 Portrait suite compatibility...............................................................................................18

Chapter 5: End of life and deprecation...................................................................19 End-of-life features............................................................................................................20 Deprecated features..........................................................................................................20

Chapter 6: Known issues.........................................................................................21 Known issues....................................................................................................................22

Scorecard Builder..............................................................................................................22 English language references............................................................................................22 Focus format change........................................................................................................22 3D display issues on virtualized Windows installs........................................................23

Chapter 7: Fixed issues............................................................................................25 Issues resolved in Portrait Miner 7.0A............................................................................26

Chapter 8: Future release plans..............................................................................27 Future release plan............................................................................................................28

4

Portrait Miner 7.0A

Introduction

In this section: • Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 • Intended audience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 • Related documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . .6

1

Purpose

Purpose This document describes what has changed in Portrait Miner 7.0A compared to its previous release (6.3A), including: • • • • •

An overview of new features and enhancements A list of known issues A list of fixed issues since the last release End of life and deprecated items Future release plans

Note: This document includes a brief list of release-specific installation considerations, such as supported platforms, languages, and application integration. For a complete step-by-step guide on installing or upgrading Portrait Miner, see the Portrait Miner Administration Guide.

Intended audience The intended audience for this document is: • System administrators • Business analysts • Customer insight analysts

Related documentation Document

Location

Portrait Miner Administration Guide

Server: \server\qs7.0A\documentation\en_portrait_miner_administration_guide.pdf

Portrait Miner Online Help

Server: \server\qs7.0A\documentation\en_miner_online_help.chm Client: \client\help\en_miner_online_help.chm

Portrait Miner DBM Reference Guide

Server: \server\qs7.0A\documentation\en_portrait_miner_dbm_user_guide.pdf Client: \client\help

6

Portrait Miner 7.0A

Chapter 1: Introduction Document

Location

Portrait Miner DBC Guide

Server: \server\qs7.0A\documentation\en_portrait_miner_dbc_user_guide.pdf Client: \client\help

Rule Miner Documentation

Server: \ext\qsruleminer\docs\apriori_docs\apriori.html

Tip: All documents are also available on Portrait Software's support website: http://support.portraitsoftware.com

Release Notes

7

Release overview

In this section: • Product overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 • Highlights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10

2

Product overview

Product overview Portrait Miner is a powerful predictive analytics solution that enables customer insight professionals to gain a clear picture of their customers for the purpose of predicting future behavior. Portrait Miner utilizes powerful 3D data visualization and rapid modeling automation to uncover important data relationships and deliver propensity scores at the push of a button, boosting predictive model accuracy and increasing the speed of analytic results. Portrait Miner can be used to predict profit-impacting behaviors and propensities, including: • • • •

customer churn cross sell and up sell opportunities campaign planning and segmentation customer satisfaction and loyalty value

Highlights • Scorecard builder algorithmic enhancements: Final rescaling for logistic regression has been removed to allow Portrait results to be compared with other standard techniques. • Portrait Uplift 7.0A: Portrait Uplift is now installed as part of Portrait Miner and controlled via licensing options. • Unicode support for Portrait Uplift: Unicode string data is now supported throughout the Portrait suite. Portrait Miner and Uplift can handle Unicode data in both field name and data. • Version 7.0 supports 5 international languages: Supported languages include: JP, DE, ES, FR and PT • New platform support: Windows Server 2012 is supported as a server platform, and SQL Server 2012 as a database platform. • End of Life announcement for Solaris: Solaris is no longer supported as a Portrait Miner or Uplift server platform.

10

Portrait Miner 7.0A

Release details

In this section: • • • •

Scorecard builder changes . . . . . . . . . . . . . . . . . . . . . . . .12 Support for Unicode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12 Regular expression changes . . . . . . . . . . . . . . . . . . . . . . .13 Installation changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13

3

Scorecard builder changes

Scorecard builder changes Within Scorecard Builder (and Scorecard Wizard by extension) the most common standard quality measure to use is logistic regression as this allows for comparison of the results with those from other statistical packages. To ensure this is a like-for-like comparison, Portrait Miner version 7.0 removes any scaling in the logistic transformation (the scaling fixes the relationship between score and probability). This makes the algorithm more standard, and has better calibration properties (for example, the mean of the predicted probability is now close to the mean of the objective). In particular for binary objectives the logistic quality measure uses the natural scaling inherent in the logistic regression (rather than adding a separate post modelling scaling stage which fits a linear transformation of the scores to give the best logistic fit to outcomes with target scaling – defaulting to even odds at 500, doubling every 20; this secondary fit is based on binning of the observed outcomes which makes it difficult to recover the original model). Model performance is still measured by Gini and its definition unchanged. For binary objectives we recommend you use logistic regression in preference to probit regression for the purposes of predictive modeling (linear regression is no longer available for binary objectives and while probit is still available as an option, future releases are likely to remove them). For continuous objectives nothing changes.

Support for Unicode Initially introduced for the Japanese version of Portrait Miner in version 6.3A, Unicode (double byte) string data is now supported as standard throughout the Portrait analytics suite (including Portrait Miner, Uplift and Explorer), whether it be data read from flat file, a database or imported from other applications. Non-Unicode (latin-1) string data is still fully supported, and backward compatibility with existing datasets (foci) means no conversion process is required. Source data files that do not have an indication of their encoding (byte order mark) are read according to the standard encoding for the locale of the machine. File names, field names and all focus metadata may also contain Unicode characters. The impact on performance due to Unicode support is minimal due to the fact that focus data is compressed on disk. However an increase in memory footprint is to be expected and where memory tuning parameters have been set in build scripts, you should consider increasing these by 25-50% where string type data is analyzed. The exact impact depends on the mix of field types involved in an operation but in general overall impact on performance is very low. While character encoding when using the command line tools is in line with the machine locale, it may be necessary to explicitly set encoding and font in a Windows command prompt. In particular, qsexportflat includes a -encoding parameter to allow you to encode the resulting flat file. Supported encodings are LATIN1, UTF-8, UTF-8-No-BOM, UTF16-LE, Shift-JIS and EUC-JP.

12

Portrait Miner 7.0A

Chapter 3: Release details

Regular expression changes As part of the move to upgrade the internal libraries for wide character data, a new regular expression library has been introduced to allow users deriving fields in the Field Derivation Language (FDL) using a standard regular expression syntax. The main changes are through use of 5 additional special characters: |?+() which require a preceding \ to be used literally (in addition to the usual .^$\[* .) Examples: Description

Old version

New version

Match postcodes which start with the letters A or B

match(“^\(A\|B\)”, Postcode)

match(“^(A|B)”, Postcode)

Match license figures with at least one 0 in them

match("0\+",License)

match("0+",License)

Remove all occurrences of $ and , from a string field

replaceall("\(\$\|,\)" , "", License)

replaceall("(\$|,)" , "", License)

Installation changes With the release of version 7.0A, Portrait Uplift is installed with Portrait Miner - so no longer requires a separate install. Note: To access Portrait Uplift, you still need a valid license to separately build uplift models. To align the connector between Portrait Miner and the rest of the Portrait suite – the MAS extension component has been removed and replaced with a single Portrait Shared Services Connector (PSS Connector) which is shipped with any release which makes use of Portrait HQ (shared between Portrait Explorer, Portrait Dialogue and Interaction Optimizer). Command line utilities which manage this connection have been renamed from cms* to pss* (for example, cmsrulepublish is now pssrulepublish). Additionally, the pss.config file contains a configuration parameter of the Portrait Shared Services' URL - if this is not set then the PSS Connector functionality is unavailable.

Release Notes

13

Installation considerations

In this section: • • • •

Upgrading to Portrait Miner 7.0A . . . . . . . . . . . . . . . . . . . .16 Supported platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16 Supported languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 Portrait suite compatibility . . . . . . . . . . . . . . . . . . . . . . . . .18

4

Upgrading to Portrait Miner 7.0A

Upgrading to Portrait Miner 7.0A When upgrading to Portrait Miner 7.0A, take note of the following notes and restrictions: • The focus format has changed with this release with the following impacts: • Focus files created or saved by version 7.0A are not backward compatible with previous versions. • Focus files created in version 6.2A or preceding versions can be read by the new version • Focus files created in version 6.3A are not compatible with the latest version. In order to make use of the files, a data migration build script has been provided in the server installation directory: ext\migrate. • The regular expression languages may require you to revisit your derivations. The introduction of new special characters |?+() typically means that when they were used literally previously – you should remove the preceding backslash (because it is meant to be special) and when there was previously no preceding backslash for these characters you should add one (because its meant to be literal). • The introduction of Unicode storage for string data means additional memory usage which you may need to take into account with any memory settings in your build scripts. We typically find increasing the certain build script steps that manipulate these string fields by between 25-50% enables your scripts to continue to work with the new formats. • If you publish predictive models for use in Portrait Dialogue or Interaction Optimizer, you should use the new pss* variants of the command line utilities. • Support for Solaris has been dropped in this release. All upgrade information relates to Windows server implementations and you should upgrade your installation to a Windows server to continue to track Portrait Miner releases. • Uninstall the previous client prior to installing the Portrait Miner 7.0A client. • Install Portrait Miner 7.0A server to a different location on the server machine. • For all Oracle databases, install the Oracle Instant Client driver (version 11.2 on Windows (32 or 64bit).

Supported platforms Server/PC

Description

Analysis servers

Portrait Miner is predominantly a client-server application, however you can opt to install Portrait Miner on a Windows desktop as a standalone application. If you go for the client-server option, the Portrait Miner analysis servers are supported on the following platforms: • Intel x86 and AMD64-based server systems running Microsoft Windows Server: • 2008 R2 SP1 • 2012

16

Portrait Miner 7.0A

Chapter 4: Installation considerations Server/PC

Description All in either 32 or 64-bit versions. Portrait Miner server runs as a 64-bit application on 64-bit operating systems and we recommend 64-bit for large scale analysis.

Client machines

The client end of the Portrait Miner supports user interaction with the application from the user's desktop. The client platforms supported are Intel-based PCs running: • Microsoft Windows Vista SP2 (including 64-bit) • Microsoft Windows 7 (including 64-bit) The Portrait Miner client runs as a 32-bit application on these platforms. The Desktop version of Portrait Miner is supported on all of these platforms and runs as a 32-bit application.

Database servers For data extract and write-back, Portrait Miner supports efficient ODBC interfaces between the analysis server and an organization's RDBMS, data mart or data warehouse. The following databases are supported, typically hosted on a separate database server machine: • Microsoft SQL Server: version 2008 R2 SP1 and 2012 are supported, hosted on Windows servers. SQL Server connections are supported from Windows 2008/2012 Servers, Portrait Miner analysis servers with a suitable ODBC driver. • Oracle: Portrait recommends Oracle Instant Client driver (version 11.2 on Windows) which provides connectivity to Oracle 10g and 11g, hosted on any Oracle supported platform. Oracle connections are supported from any Portrait Miner server platform. • NCR Teradata: Portrait recommends the Teradata TTU13.0 drivers are used with Portrait Miner 7.0A which support Teradata versions V2R6.2 and above hosted on any NCR Teradata supported platform. Teradata connections are supported from any Portrait Miner server platform. • ODBC: version 3.51 supported from all Portrait Miner server platforms to databases with appropriate native client ODBC drivers. Suitable ODBC drivers for all currently supported database platforms are available from the database vendor including Sybase ASE and IBM DB2 UDB. • Statistical Datasets: direct access to the following data formats: all versions of dBASE and compatible formats, including Microsoft Excel (reading all versions through Excel 2007 and writing Version 2.1 files), SAS (SAS 6 for Windows, common format for all platforms; uncompressed data only), SAS Transport, SPSS for Windows (compressed and uncompressed files), SPSS Portable. These data formats are available from all Portrait Miner analysis servers. • Flat File: extensive support for importing and exporting data in ASCII flat file format (delimited and fixed-width) is provided. 32 -bit database drivers are required for database connections from Portrait Miner Desktop installations (which runs as a 32-bit application on all platforms).

Release Notes

17

Supported languages

Supported languages Portrait Miner 7.0A is available in English, French, German, Japanese, Spanish and Portuguese (Brazilian).

Portrait suite compatibility Portrait Miner 7.0A is compatible with the following Portrait products: • • • • • •

Portrait Uplift 7.0A Interaction Optimizer 6.0 Portrait Dialogue 6.0 Portrait Optimizer 7.0A Portrait Explorer 4.0 Portrait HQ 2.0

Note: Self-Service Analytics 2.2A is only compatible with Quadstone System 5.3B.

18

Portrait Miner 7.0A

End of life and deprecation

In this section: • End-of-life features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20 • Deprecated features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20

5

End-of-life features

End-of-life features The following components are now end-of-life and no longer supported: • Portrait Miner 5.x. Portrait's support policy includes support for two major releases (v6.x and v7.0A). Customers on version 5.x or earlier should contact Portrait Support for guidance on how to upgrade.

Deprecated features This release deprecates a number of components, which are all still supported in the Portrait Miner 7.0A, but support for which will be removed from the product in a future release. These include: • • • •

20

Windows Server 2008 SP2 as a server platform. Windows Vista as a client platform. SQL Server 2008 SP3 as a Portrait Miner database platform. Oracle 10g R2 as a Portrait Miner database platform.

Portrait Miner 7.0A

Known issues

In this section: • • • • •

Known issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22 Scorecard Builder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22 English language references . . . . . . . . . . . . . . . . . . . . . . .22 Focus format change . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22 3D display issues on virtualized Windows installs . . . . .23

6

Known issues

Known issues A number of known issues in Portrait Miner 7.0A have been highlighted below.

Scorecard Builder Known Scorecard Builder issues include: • Probit regression should not be available to Scorecard Builder and linear regression should not be available to Scorecard Wizard • Include/exclude displays Gini but selects variables on the basis of the current quality measure – it should report change on the same measure as it selects variables. • Stepwise linear fit should be removed from auto Include/exclude option • Stepwise quality measure should use the quality measure throughout (and not build a trial scorecard with linear regression) • Weights of evidence (WoE) measures for empty bins can be non-zero due to rescaling of the original WoE values.

English language references In translated versions of Portrait Miner there are still some English language variants visible in the application interface. This is due to: • the English phrase being the most appropriate to use • new tool development, for example, Rule Miner which has not been localized yet • the user interface mistakenly not being translated (rare).

Focus format change The focus file version has been updated with the release of Portrait Miner 7.0A, and is no longer compatible with previous versions of Portrait Miner.

22

Portrait Miner 7.0A

Chapter 6: Known issues

3D display issues on virtualized Windows installs The libraries used to implement 3D views in Portrait Miner have been upgraded. On some platforms virtualization of Windows does not provide the necessary support for 3D visualization in this version of Miner. It is possible to revert to the previous 3D libraries if this issue arises. Please contact support for advice if this configuration is required.

Release Notes

23

Fixed issues

In this section: • Issues resolved in Portrait Miner 7.0A . . . . . . . . . . . . . . .26

7

Issues resolved in Portrait Miner 7.0A

Issues resolved in Portrait Miner 7.0A The following issues raised in Portrait Miner 6.3A have been fixed in version 7.0A.

26

Bug/issue

Type

Description

Internal

Bug fix

qsjoin fails on out of sequence error depending on memory settings. This issue arises when filters are applied to the .ftr and a memory option given. Fixed.

Internal

Bug fix

The extension component excelmetadata which allows preparation of Focus metadata using MS Excel is not available in Miner 6.3. Fixed.

Customer

Bug fix

After deriving a string(0) empty string "" field, a system crash occurs 38-3-50-42-3-124-1143-3. Fixed.

Customer

Bug fix

The average predicted probability is not exactly equal to average outcome rate. Fixed.

Internal

Bug fix

qsderive takes a long time writing data after deriving a long string field. Fixed.

Portrait Miner 7.0A

Future release plans

In this section: • Future release plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28

8

Future release plan

Future release plan To aid your planning, we are scheduling a further Portrait Miner release during 2013. Beyond this next release timescale we are considering the following areas of enhancement. These are provided here purely for information and should only be used for the most general of planning purposes. • Interactive aggregation and linking of relational data to simplify the preparation, exploration and discovery process for hierarchical data (e.g. household, individual, account, transaction, event) • Spatial analytics FDL extensions allowing you to calculate distance and angle between latitude and longitude pairs, and geo-lookup using latitude and longitude to reference k-th nearest item in a reference table. • Spatial analytics data-preparation extensions allowing you to perform geocoding, drive-time and distance calculations, geodemographic appends and MapInfo data import and export. • Enhanced integration with Microsoft Office and SQL Server Reporting Services, making it easier to present compelling and configurable results from your analysis. • Improved analysis of geographical data in Map Viewer, with better integration to MapInfo. • Customer “baseball cards” to provide rapid visual insight and validation for individual customer records. • Improved user experience for interactive modeling in Scorecard Builder and Cluster Builder. • Comprehensive support for case weighting, to correct for biased datasets in reporting, visualization and modeling tools. • Further enhanced deployment of results to operational systems, making it easier to transfer models, scores and lists directly into systems such as Portrait Dialogue, Portrait Interaction Optimizer, PMML, SAS, and RDBMS. • Interoperable focus format: allow third party tools to read and write the Portrait focus format. • Broader support for model comparison, monitoring and management, including pre-integrated model tracking, champion-challenger comparison, performance and exception alerts, and automated model rebuild as required. • Extended platform support including Windows 8 and Linux as Portrait Miner server platforms; Windows 8 as a client and desktop platform; and Oracle 12 and Teradata V2R7 as database platforms.

28

Portrait Miner 7.0A