Tag: DataScience

Evaluating Machine Learning models when dealing with imbalanced classes

Evaluating Machine Learning models when dealing with imbalanced classes

By | | Comments 1 comment

In this blog post I talk through an example of how to pick the best model when you deal with these kind of problems. I also touch the subject of cost-sensitive predictions, introducing some code to generate plots that will help you understand your model in cost fashion. Even more important, it will be essential for grasping the full business impact when moving to a data driven world!

#DataScience #R #MachineLearning #AzureML




Evaluating Machine Learning models when dealing with imbalanced classes – Developing Analytics Solutions with the Data Insights Global Practice – Site Home – MSDN Blogs

Sander Timmer, PhD. In real-world Machine Learning scenarios, especially those driven by IoT that are constantly generating data, a common problem is having an imbalanced dataset. This means, we have far more data representing one outcome class than the other. For example, when doing predictive …

Check this out on Google+ 3 1

Using Azure Machine Learning Notebooks for Quality Control of Automated Predictive Pipelines

Using Azure Machine Learning Notebooks for Quality Control of Automated Predictive Pipelines

By | | Comments 0 Comment

When building an automated predictive pipeline, to have periodically batch-wise score new data, there is a need to control for quality of the predictions. The Azure Data Factory (ADF) pipeline will help you ensure that your whole data set gets scored. However, this is not taking into consideration that data can change over time. For example, when predicting churn changes in your website or service offerings could change customer behavior in such a way that retraining of the original model is needed. In this blog post I show how you can use #Jupyter Notebooks in +Microsoft Azure Machine Learning (AML) to get a more systematic view on the (predictive) performance of your automated predictive pipelines.

#DataScience #MachineLearning #Azure #AzureDataFactory #Python #Notebook




Using Azure Machine Learning Notebooks for Quality Control of Automated Predictive Pipelines – Developing Analytics Solutions with the Data Insights Global Practice – Site Home – MSDN Blogs

By Sander Timmer, PhD, Data Scientist. When building an automated predictive pipeline to have periodically batch-wise score new data there is a need to control for quality of the predictions. The Azure Data Factory (ADF) pipeline will help you ensure that your whole data set gets scored.

Check this out on Google+ 1 1