Last night, our Oracle Analytics instance was upgraded to 5.9.
It was long awaited upgrade, which introduces some nice new features and improvements. My personal interest lies with the enhanced Machine Learning support, however, there are several other features that might draw your attention.
Let me briefly talk about some of them.
Additional functions in Database Analytics step in Data Flows
Two new Database Analytics steps in Data Flows are:
- Frequent Itemsets and
- Text Tokenisation.
About Frequent Itemsets, I'm describing in detail in my separate blog Oracle Analytics 5.9: Frequent Itemsets, so let me not repeat myself here.
Text Tokenisation allows users to analyse textual data by breaking it down into distinct words and counting the occurrences of each word.
Technically, when data flow is run, a new database table would be created, table is named DR$IndexName$I, which contains the token text and the token count related details. Once created, this table can be used to create a new dataset and used in data visualisation:
Machine Learning Improvements
Frequent Itemsets are one of the improvements which can be also treated as Machine Learning Improvements. For complete Market Basket Analysis, Frequent Itemsets are probably not enough, as it would be expected to have options to analyse recommendations too. But I firmly believe, Oracle Analytics will get this in future releases.
The second improvement relates to registered Oracle Machine Learning (OML) models that reside in Oracle Database and are registered with Oracle Analytics.
When OML model is created in Oracle Database, then a set of database tables and views, so called model detail views, are created. These model detail views provide detailed information about machine learning models so any user can assess their quality and check any other model details.
Up until now, when you registered OML model in Oracle Analytics, you didn't have any detailed information about registered OML models. Using other words, you had to trust machine learning model creator regarding its quality.
In Oracle Analytics 5.9, when OML model is registered, then users can check which model detail views are related to machine learning model registered. These model detail views can now be used, imported as datasets and can be analysed.
More detailed explanation of model detail views in my blog post Oracle Analytics 5.9: Model detail views for registered OML models.
Support for Web Map Service and Tiled Maps
Major improvement has been made in relation to the background maps that can be used when analysing data on maps. Support for very popular Web Map Services and Tiled Map Services has been added in order for these map types to be used as background maps.
I am writing about WMS and XYZ maps in a separate blog Oracle Analytics 5.9: Using Web Map Service (WMS) and Tiled Maps (XYZ). Please feel free to check for more details there.
Data Visualisation Improvements
- Look & Feel in general has been optimised.
For example, left bar with Elements, Visualisations and Analytics icons has been moved to horizontal position under the blue bar. This freed up space for the visualisations which can stretch, if all panels are minimized, from left to right edge of the screen.
Additionally, you will observe that progress bar for running queries are now slightly different. A new progress circle that can be found on the right top corner indicates running queries.
In Data Flows, steps on the left are now smaller and, at least for me, easier to use.
- Sorting by multiple columns. Previously, it was not possible to sort tables, pivot tables, graphs, by more than one column. With 5.9, this has changed, I and believe many users will find this change quite useful.
Additionally, you can choose None, meaning data will not be sorted by that attribute. - New Area Chart visualisation. Beside stacked Area chart, now with this chart you can stack various categories one on top of each other (not above!). Area layers are now transparent, so you actually analyse data using this visualisation quite effectively.
Data Actions
- Avoid passing duplicate filters. When you create a data action using multiple data points to filter another canvas, and you return to the original canvas and invoke the data action using different data points, the new expression filter replaces the previous filter on the target canvas. In some cases this was truly annoying.
- Support single/multi selection. You can restrict the invocation of a data action to work only for a single value by setting Supports Multiple Selection to Off.
You can set this when the selection of multiple data points will result in an error (for example, with some third-party REST APIs).
More Information
For more information about Oracle Analytics, please take a look at complete list of new features and resolved bugs on Oracle Analytics documentation pages and there is also video series published on YouTube.