SAS 模型管理 发表评论(0) 编辑词条

SAS® 模型管理
概述
SAS Model Manager对冗长而且往往容易出错的分析模型创建、管理和部署步骤进行简化，并持续对它们的准确性和有用性进行检验核对。由于分析挖掘技术在业务流程中扮演着非常重要的角色，因此，企业必须提高模型输出结果的准确度，避免对模型结果的错误解释。
产品益处
更快地对模型进行管理和生产部署。SAS Model Manager提供方便使用的图形用户界面，在对模型进行注册、测试和确认的可重复流程中为用户提供指导。可以对模型从创建到部署至实时或批量评分系统直到退役的整个过程进行监督。
通过一个整合的环境对模型性能进行跟踪和监督。循环式的组织和跟踪框架确保分析模型在整个模型生命周期过程中都能像预期的那样正常发挥功能。在对模型进行测试和比较的同时，性能基准报表自动生成。随着模型在企业内的部署，性能衡量标准也被传播至整个报告渠道。
遵守诸如萨班斯-奥克斯利法案和 BASEL II这样的法规要求。SAS Model Manager提供灵活而独特的法规遵从及合法性报表。可以通过专利的中央数据库、生命周期模板和元数据管理系统来获取最佳实践。在困难的部署分析步骤中为用户提供指导，包括从模型创建到向生产运作环境进行部署。
特色
对组织模型进行集中、安全地存储

根据项目对模型进行存储。
在项目中对冠军和挑战者模型进行单独版本的设置和维护。
对模型报告和分值代码测试所使用的必备数据源进行映射。
对所有主要行动、用户定义的备注进行事件记录。
附上支持文件，例如，Microsoft Word文档，Microsoft Excel 电子表格，HTML 文件等。

预制的标准数据挖掘模型注册模板

导入SAS Enterprise Miner模型。
导入SAS/STAT 和 Base SAS模型。
数据库元数据汇总报表，例如，模型数量，评分岗位数量。
模型老化曲线。
对每个目标变量和输入变量在所有模型中的使用频率进行统计。

按照属性对模型库进行查询

算法类型。
输入变量或目标变量。
模型创建者。
模型ID。
生命周期批准用户。
查询属性组合。
能够添加用户定义的查询键。

安全的模型存储和访问管理

备份和恢复功能。
覆盖保护。
事件记录。
用户身份认证/访问优先权管理。

在将模型导出之前对评分逻辑进行有效性确认

定义测试和生产评分工作。
将模型导出至SAS元数据库。
生产评分。
将模型更新发布至不同的评分渠道。

测试和生产生命周期过程中的模型性能和对比报告

模型性能报告：
评分数据汇总计划。
变量分布图。
特征图。
稳定度图。
提升图。
接受者操作曲线和吉尼图。
柯尔莫哥洛夫-斯米诺夫图。

模型对比报告：
模型概况报告。
变量增量报告。
动态提升。
模型监控报告。
特殊SAS代码报告编辑器。

分析模型整个生命周期管理

模型生命周期模板，用于协同项目管理。
以任务为导向的重大任务完工和审批签字。
定义开始日期和完工日期。
流程进度状态报告。
系统要求
客户端环境
Windows (x86-32): Windows XP Professional, Windows 2000 Professional
服务器环境
AIX Release 5.1, 5.2, 5.3 on POWER
Solaris on SPARC: 版本 8, 9 和 10
Windows (x86-32): Windows 2000 Server, Windows Server 2003
必要/不包括的软件
SAS Enterprise Model Management 2.1 是一个打包销售产品，包含Base SAS, SAS/STAT® and SAS Enterprise MinerTM software.
SAS Model Manager 2.1 requires Base SAS 和 SAS/STAT 软件。
中间层组件
两个打包产品都包括SAS Analytics Platform 和 Xythos WebFile Server with PostgreSQL。
其它客户端组件
客户端上的SAS Management Console（包含在Base SAS内）：
AIX Release 5.1, 5.2, 5.3 on POWER
HP-UX PA-RISC: Release 11i+ Version 1, 2 and 3
HP-UX Itanium: Release 11i+ Version 1, 2 and 3
Linux for x86 (x86-32): Red Hat Linux 8.0, RHAS 2.1, RHEL 3.0 和 4.0, SuSE SLES 8 和 9
Linux for Itanium (64-bit): Red Hat RHEL 3.0
Solaris on SPARC: 版本 8, 9 和 10
Windows (x86-32): Windows 2000 Server, Windows Server 2003
Windows (64-bit on Itanium): Windows Server 2003
SAS Model Manager, with a patented, secure model repository complemented by a rich metadata structure and project templates, streamlines the tedious and often error-prone steps of creating, managing and deploying analytical models. It enables organizations to effectively create, manage and deploy statistical, predictive, classification and analytical scoring models in an enterprise computing environment.

Benefits
Reduces time to manage and deploy models into production.
Provides an integrated environment for tracking and monitoring model performance.
Enables compliance with regulatory requirements.
Read more

Features
Central, secure repository for organizing models
Validate the scoring logic before exporting models to production
Monitoring and reporting on model performance during test and production life cycles
Overall lifecycle management of analytical models
Read more
Screenshots
SAS Model Manager's easy-to-use GUI.

More ScreenshotsEnlarge
How SAS® Is Different
SAS Model Manager enables organizations to effectively create, manage and deploy statistical, predictive, classification and analytical scoring models in an enterprise computing environment.
It provides a patented, secure, centralized repository for storing and organizing models backed by extensive documentation templates for each model.
Accountability metrics and validation of analytical steps, from the time of creation through deployment into real-time or batch scoring systems, continues until the time a model must be retired.
Benefits
Reduces time to manage and deploy models into production. SAS Model Manager provides an easy-to-use graphical user interface that guides users through a repeatable process for registering, testing and validating models. Accountability metrics and version control status reports track who changes what, when control is passed from one area to another, etc. Models can be monitored from their creation to deployment into real-time or batch scoring systems until they are retired.
Provides an integrated environment for tracking and monitoring model performance. With its iterative framework, SAS Model Manager ensures analytical models are functionally performing as intended throughout the model life cycle. As models are tested and compared, performance benchmarking reports are generated. As they are deployed, performance metrics are pushed over established reporting channels. Modelers can easily collaborate and reuse models, and automated alerts can be set to detect when the scoring results are changing over time indicating model decay.
Enables compliance with regulatory requirements. SAS Model Manager’s flexible and unique compliance and validation reporting are highly sought-after by those facing increasing regulatory requirements. Valuable best practices can be captured via the patented centralized data repository, lifecycle templates and metadata management system. Users are guided through the difficult steps of deploying analytics from creation into the production operational environment.
Features
Central, secure repository for organizing models
Project-based storage of models.
Set up and maintain separate versions of champion and challenger models within a project:
Freeze a version.
Set a default version for the project.
Champion challenger model promotion.
Map prerequisite data sources used for model reporting and score code testing:
Training and test tables.
Score input and output tables.
Performance tables.
Project input and output tables.
Accounting and auditability:
Event logging of all major actions.
User-defined notes.
Attach documents (Microsoft Word documents, Microsoft Excel spreadsheets, HTML files, etc.) and add version control.
Prebuilt templates for registering standard data mining models:
Prediction.
Segmentation.
Classification.
Scorecards.
User-defined template.
GUI or SAS macro model registration.
Model Import Template editor.
Optional SAS macro model registration support.
Import multiple SAS Enterprise Miner models:
General properties such as model name, type of algorithm, creation date, modification date, etc.
Model inputs required for scoring.
Model outputs generated by scoring.
Score code including preliminary transformations.
Associated scoring tasks.
Advanced view of the SAS Enterprise Miner process flow diagram.
Import multiple SAS/STAT and Base SAS models:
Training code.
Score logic.
Estimate tables.
Target and input variable.
Import and export PMML model code with inputs and outputs.
Repository metadata summary report, such as:
Number of models; number of scoring jobs.
Model aging profiles.
Frequency counts of how often each target and input variable has been used across the model portfolio.
Query the model repository by attributes, such as:
Type of algorithm.
An input or target variable.
Model creator.
Model ID.
Lifecycle assignee or approval user.
Combination of query attributes.
Ability to add user-defined query keys.
Secure, reliable model storage and access administration:
Backup and restore capabilities.
Overwrite protection.
Event logging.
User authentication/access privilege administration.
Validate the scoring logic before exporting models to production
Define test and production score jobs:
Map required inputs and outputs.
Add pre- and post-SAS code.
View log and results table.
Create interactive graphs.
Export models to SAS Metadata Repository.
Production scoring:
Mining Results Transformation available in SAS Data Integration Studio.
Model Scoring Task available in SAS® Enterprise Guide®.
Publish models directly to SAS Real-Time Decision Manager.
Publish model updates to different scoring channels:
E-mail notification sent to subscribers.
Store results to a file system or post to a corporate intranet.
In-database model deployment:
Integrated with the SAS Scoring Accelerator for Teradata to publish and validate Teradata Scoring functions for native scoring in Teradata Database.
Monitoring and reporting on model performance during test and production life cycles
Model Performance reports:
Programs for summarizing scored data.
Variable distribution plots.
Characteristic chart.
Stability chart.
Lift chart.
Receiver Operating Curve and Gini charts.
Kolmogorov-Smirnov chart.
Model Comparison reports:
Model profile report.
Delta report.
Dynamic lift.
Model monitoring report.
Ad hoc SAS code report editor.
HTML, RTF, PDF and Microsoft Excel output formats.
GUI for creating performance monitoring reports.
Overall lifecycle management of analytical models
Model lifecycle templates for collaborative project management:
Basic, Standard, Extended and User defined.
Model lifecycle template editor for User defined.
Task-oriented milestone completion and approval signoff.
Define start and completion dates.
Progress completion status reports.
Screenshots
Build more models faster with SAS Model Manager’s easy-to-use GUI.

Enlarge

System Requirements
Client environment

Microsoft Windows (x86-32): Windows XP Professional, Windows Vista*, Windows Server 2003 family
Microsoft Windows on x64 (EM64T/AMD64): Windows XP Professional for x64, Windows Vista* for x64, Windows Server 2003 for x64
Server environment

AIX: Version 5.3 and Version 6.1 on POWER architectures
HP-UX Itanium: HP-UX 11iv2 (11.23), 11iv3 (11.31)
Linux for x64 (EM64T/AMD64): RHEL 4 and 5, SuSE SLES 9 and 10
Solaris on SPARC: Version 9, 10
Microsoft Windows on x64 (EM64T/AMD64): Windows XP Professional for x64, Windows Vista* for x64, Windows Server 2003 for x64
* NOTE: Windows Vista Editions that are supported include Enterprise, Business and Ultimate

Required/not included software

SAS Model Manager requires
Base SAS and SAS/STAT® software.
SAS Enterprise Model Management is an inclusive sales bundle containing Base SAS, SAS/STAT and
SAS® Enterprise Miner™ software.