Advanced analytics in Oracle Autonomous Database

 In a previous post we talked about the fact that the new free tier Oracle had been announced,including Oracle Autonomous Database.In this post we will focus on showing Oracle Autonomous Database notebooks functionality. Up to now, notebooks was oriented to the researchers and academic world, in general. But everything is changing, and now the companies have started to interesting in take benefit of his data and turn to data-driven companies. Companies want to generate new types of analysis, not only reactive, also predictive analytics.

 

Autonomous Database notebooks are based on Apache Zeppelin technology, with the peculiarity that Oracle offers us the environment configured and integrated with the database, so the code will be executed in the database engine, giving us a better performance and eliminating data movements between different applications.

 

 

Notebooks are work environments where developers and data scientists could develop advanced algorithms, share their work, schedule jobs and present their work in a visual and documented way.

 

The capabilities of workbooks in business are unlimited. Let’s see an example of how they could manipulate a dataset with this tool and ease of use.

 

The first step is to create a database table. For this purpose, it has been used an open data source, but any data source can be used. This entity stores last years’ sales of an office supplies store. For the table, two fields are necessary at least: datetime field and a numeric field representing a fact, in this case, sales quantity.

 

Once the notebook is created, the next step is select all data in a table format. It’s only to check visually the data to work:

 

 

 

We want to know the sales for the following 4 quarters, so a forecasting algorithm fits for this purpose perfectly. Autonomous Database offers us different advanced algorithms, so let’s use those features to create our predictive model. This code removes the models, creates a view with time and sales values we want to use in the model and finally, creates a table to store all parameters applied to the algorithm. Once the configuration is finished, let’s create de model.

 

 

In the following steps we are going to check the model configuration. Those steps are given only for information and are not necessary. We have included it to give more information how Oracle analytics packages works.

 

It’s recommended to review Oracle documentation to understand all attributes and options due the large quantity of functionalities, configurations and complexity. It’s time to check how the model has worked and the result obtained. The models have been configured to group the sales by quarter, so the forecast will be by quarter.

 

The model generates three values: forecast and the lower and upper limits or confidence intervals. Finally, shows graphically the result values:

 

 

As we have seen, Oracle Autonomous Database notebooks are a powerful tool for developers or data scientist to develop models and advanced calculations, using the analytics packages this database offers. The model results are persisted in the database, now it’s possible to use all this data to create visualizations, analysis or merge with other data sources using tools like Oracle Analytics Cloud or another third party reporting tools. On the other hand, those models could be scheduled periodically to have always “fresh-data” for our reports or analysis.

Twitter
LinkedIn
Evolution, innovation and transformation
5 expertise + 42 specializations endorsed by Oracle
Our value proposition
100% Oracle posts
Follow our day-to-day activities