Welcome to the latest One Model product release update. This article provides an overview of the product innovations and improvements to be delivered on 31 May 2023.
One AI
Download Data Sets from Recipes
The ability to download both the predict and train/test datasets in their entirety from Recipes has been added in this release. This allows you to examine the dataset in its entirety and/or leverage the data in external data science tools. This new capability further cements One Model’s position as the world’s most ethical, open and transparent ML/AI/Predictive platform for People Analytics.
The dataset download feature can be found by expanding the Do you want to download the dataset generated by this recipe? question in a Recipe. The two files are available separately and are comma delimited (.csv) formatted. (ref 2283)
NULL Filling for Generative Attributes
To prevent predictive Generative Attributes from being dropped by One AI prior to being considered as features, we have added an easy-to-use null filling option. This option is available on the Generative Attributes creation or editing screens in One AI Recipes. When selected, this option replaces any NULL values for that Generative Attribute with a 0.
The primary reason Generative Attributes are dropped as features is missing data, or NULL values. Using a Generative Attribute of Promotions as an example, those employees who have never received a Promotion receive a NULL value for that attribute by default. Selecting the Fill NULLs with 0 option replaces those missing values with 0s.
Please note that this option should only be used for Generative Attributes that are sums or counts. It would not be appropriate to fill NULLs with 0s for averages. More advanced NULL filling is available as a Per Column Intervention in the Augmentation Configuration and can be read about in the Refining a Machine Learning Model in One AI help article. (ref 2130)
Positive Label Selection
For binary classification models, there are always two labels (possible values for the thing you’re predicting). While predictions are made for each label, the results are easier to interpret when you focus on one of the labels. This is “the positive label”. Using binary attrition risk as an example, the labels might be “Terminated” and “Did Not Terminate”. Since we want to focus on the risk of terminating, “Terminated” is the positive label.
In this release we added the ability to select the positive label in Recipes rather than relying on One AI to select it for you.
If you leave Let One AI decide selected, the maximum sorted value will be chosen.
After running a binary classification augmentation, the positive label will be listed in the Prediction Details section of the resulting Modeling Report. The Feature Analysis section of the report will be focused on the positive label. (ref 2271)
New One AI Help Articles
A number of new One AI related help articles have been published to the One Model help center recently. The following are links and brief descriptions.
-
Introduction to One AI
One AI enables HR professionals to access statistical and machine learning insights alongside their people analytics data in One Model
-
Introduction to Machine Learning in One AI
Learn about machine learning in One AI
-
How to Create a Voluntary Attrition Risk Model in One AI
http://help.onemodel.co/en/articles/7939771-how-to-create-a-voluntary-attrition-risk-model-in-one-ai
A step-by-step guide for creating a predictive voluntary attrition risk model in One AI
-
Refining a Machine Learning Model in One AI
Maximize insights from One AI by refining your model
-
One AI Machine Learning Algorithms and Settings
A number of models using different algorithms and settings are compared each time a machine learning augmentation is run
One Model Help Center
- We are planning some big changes and improvements to the One Model Help Center with a major upgrade planned for the product release on 28 June 2023. You will still find the Help Center link at the top of the application navigation bar (direct link help.onemodel.co) but all of the links to help articles and categories will be different to what they are today, which is why we wanted to give you as much advance warning of the upcoming changes as possible in case you are linking to these resources from other places.
Platform Improvements
-
We made some improvements to the Storyboard Filter Bar Search capability as some customers with very large dimensions were having difficulty with their search requests. Users should find improved search processing and load times. (ref 16085).
Data Processing Improvements
-
We now keep all transformation tables created by the processing scripts by default after a new data load, which allows for better auditing of what was done for that most recent data load. (ref 15923)
Minor Improvements & Bugs Squashed
-
When editing a Storyboard you are able to link a tile to another Storyboard, but we had a bug stopping you from removing a link as the option to remove a link by choosing and saving ‘No Storyboard Link’ on a tile was not working. With this fix, users can again remove links from their Storyboard tiles by choosing and saving ‘No Storyboard Link’. (ref 16044)
-
We recently released the option to export a Storyboard chart to a csv file. That feature had an initial limitation. If you drilled down in a chart by clicking through a dimension the export will only include the top level data. We have enhanced this to now capture your drill-down context in the csv export. (ref 15756)
-
We fixed a minor issue with the display of text in the Storyboard Settings Panel when setting conditional formatting for a table. If you selected the Data Range option the text in the panel could get a little blurry. This was only visible for some users depending on their Windows display settings. (ref 15538)
Comments
0 comments
Please sign in to leave a comment.