Machine Learning - Score

Published: August 25, 2015

Updated: March 23, 2017

This section lists the modules provided in Azure Machine Learning for scoring. When you use one of these modules in your experiment, you must attach an already trained machine learning model, and some new data. The scoring module then generates scores from the new data based on the model. Such scores typically represents predictions based on the model, and can take many forms:

• A list of recommended items
• Forecasts, for time series models
• Estimates of projected demand, volume, or other numeric quantity, for regression models
• Cluster assignments
• A predicted class or outcome, for classification models

Azure Machine Learning provides different scoring modules, depending on the type of model you are using:

• Use Score Matchbox Recommender if you want to generate recommendations, find related items or users, or predict ratings.
• Use Assign Data to Clusters if you want to cluster new data based on an existing K-Means clustering model.
• Use Score Model for all other regression and classification models, as well as some anomaly detection models.

Related Tasks

• Special scoring modules are provided for Vowpal Wabbit. See Text Analytics.
• Scoring for special classes of images on pretrained models can be performed using the OpenCV Library.
• The Time Series Anomaly Detection generates scores that represent potential deviations from a trend.

Warning
Before you perform scoring on your dataset, check for missing values and nulls. When data used as input for scoring has missing values, the missing values are used as inputs. Because nulls are always propagated, the result is usually a missing value.

The following articles provide real-world examples of how you can use a machine learning model for scoring:

• Develop a predictive solution

• Survival Analysis

• Anomaly Detection

• Lexicon-based sentiment analysis

The Score category includes these modules.

Module Categories and Descriptions
Train
Score
Evaluate