转换为指标值 [英] Convert to Indicator values

问题描述

我有一个Azure ML实验.我已经使用转换为指标值模块将分类变量转换为列:类别变量是InstituteType，并且具有3个值之一:Public，Private，PrivateNonProfit.

I have an Azure ML experiment. I have used the convert to indicator values module to transform a categorical variable into columns: The category variable is InstituteType and has one of 3 values: Public,Private,PrivateNonProfit.

训练实验效果很好.我创建了预测性实验并部署了Web服务.

The training experiment works fine. I create the predictive experiment and deploy the webservice. 

当我使用Request-Response链接测试Web服务时，代码失败，并出现以下错误:

When i use the Request-Response link to test the webservice, the code fails with the following error:

{代码":"LibraryExecutionError"，消息":模块执行遇到内部库错误."，详细信息":[代码":"ScoredFeaturesMustMatchTrainingFeatures"， ; target:"得分 模型(AFx库)，消息":表:被计分的数据集必须包含训练期间使用的所有特征，缺少特征:"institutetype-PrivateNonProfit"，"institutetype-Private". }]}}

{"code":"LibraryExecutionError","message":"Module execution encountered an internal library error.","details":[{"code":"ScoredFeaturesMustMatchTrainingFeatures","target":"Score Model (AFx Library)","message":"table: The data set being scored must contain all features used during training, missing feature(s): 'institutetype-PrivateNonProfit', 'institutetype-Private'."}]}}

在我的测试输入数据中，我的值是研究所类型的公开".看来系统希望在单个测试消息中看到所有可能的值.

In my test input data, i had the value Public for Institute Type. It looks like the system expects to see all the possible values in the single test message. 

请问对此有何建议?

谢谢

推荐答案

在几个地方可以调查发生的事情

There are couple of places where you can investigage whats going on

1.您是否传递了原始数据集中用于构建模型的所有列?使用网络服务来查看您的模型是否有效.

1. Are you passing all the columns that were in your original dataset that was used for building the model? Use the web service to see if your model works.

2.您是否正在使用基于过滤器的功能选择?如果是这样，应将其关闭以进行评分.

2. Are you using filter based feature selection? If so that should be turned off for scoring.

这篇关于转换为指标值的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！