Twitter Sentiment Analysis Collection

December 4, 2015

Report Abuse
This collection demonstrates how to create sentiment analysis training & predictive experiments and publish as a web service
This collection has a set of two experiments: - [Training Experiment](http://gallery.azureml.net/Details/0de07eec19d54fd2aecf98bf330323b3) - [Predictive Experiment](https://gallery.cortanaanalytics.com/Experiment/5ec8a0f4a6bb499088f11fb450e468ed) This collection shows how to overcome the challenge where "filter based feature selection" module changes the feature set when the module runs in training and predictive experiment. In this collection, we use the module "select column transform" to persist the transformation from training to predictive experiment to ensure that the same set of features are used across training and predictive experiments. In training experiment, we add an additional module "Select Column Transform". Next, we connect the output of "Filter based feature selection" to this new module "Select Column Transform". Now run the training experiment. Once run is finished, save the output of "Train Module" as a new trained model and "Select Column Transform" as a new transform. More details are available in respective experiments