Machine Learning in the Cloud, without any panic!

Don’t panic! The Official Hitchhiker’s Guide to Oracle’s Machine Learning and AI in the Cloud

So you want your database in the Cloud to have the ability to carry out Machine Learning, easily and economically… but you are worried you might have to revisit some graduate maths courses to remember anything other than rudimentary statistics and understanding of algorithms.

Well relax… and certainly – in the words of Hitchhiker’s Guide to the Galaxy author Douglas Adams – Don’t Panic! That’s the advice of Miracle Finland CEO Heli Helskyaho, who gave a presentation on ‘Machine Learning in the Cloud – Without Any Panic’ to the Data Vault User Group. Heli stressed that whilst she was using Oracle tools for this presentation – other equivalent tools are available from different vendors.

Oracle leaves you with the job of finding the data and preparing it so that it is usable for Machine Learning. the Oracle tools take care of the rest! And the good news is the service is still at a point where there is more to come in terms of Oracle’s capability regarding forecasting. Heli, who graduated from the University of Helsinki with an MSc in computer science, has worked with Oracle since 1993, having joined the IT industry in 1990. The lecturer and researcher, who is currently working on her doctorate in multi-model and converged databases,has written a number of books, including Oracle SQL Developer Data Modeler for Database Design Mastery and Machine Learning for Oracle Database Professionals.  

Preparing data and scoring model before launch

  The certified Data Vault 2.0 practitioner was helped by her son Matias, who has been working in data analyticswith experience in the Internet of Things (IoT) and Machine Learning.

Heli’s presentation started off by explaining the Machine Learning process. It starts by defining the task and understanding the task, before collecting the data and understanding it. From the data attributes,features and columns – the data is prepared and transformed while the data models are created and evaluated. Then it is a question of scoring the model before its deployment, with both the data and the model being monitored after it goes live. In other words… after exploratory data analysis and data visualisation, feature selection and engineering is followed by algorithm selection and featured encoding before hyper-parameter tuning and model evaluation, ending with model interpretation and the ability to explain it. The strength of Oracle is that you can take advantage of its Auto Machine Learning capability – meaning you don’t have to have the detailed knowledge, or do the work yourself.  

Take advantage of Oracle’s ‘Always Free Services’

  Even better, Oracle Cloud provides a free tier to allow you to build, test and deploy applications,allowing you to use the ‘Always Free Services’ to create the shortcuts in Auto Machine Learning and AI by creating an autonomous database. Oracle APEX is a low-code development platform which will help you build a scalable and secure enterprise application, deployable anywhere. The web-based interface enables you to perform data analytics, data discovery and data visualisations. Heli and Matias walked through the process of using Oracle Machine Learning User administration – the user must have the privileges to OML. They explained the database actions that allow you to execute queries and scripts and create database objects. The data modeller creates the relational diagrams for database objects while Liquibase tracks schema changes.  

Benefits of Oracle Auto Machine Learning

  Finally, Oracle Machine Learning offers several components to make your analytics set-up much more effective, and offer a better financial return. Oracle’s data tools will import and export data quickly, and load or access data from local files or remote databases. It will also create a data catalogue to understand data dependencies and the impact of changes as well as discovering anomalies, outliers and hidden patterns in your data. They also allow you to create visualisations such as area, bar and pie charts plus other popular charting methods, while an Excel plug-in allows you to create and edit custom queries and see the results in a worksheet. From there the fun can begin with AutoML, creating and running experiments, managing and deploying ML models, powered by 10 standard algorithms. Oracle also allows you to trade off greater faster results – or greater accuracy.

The Support Vector Machine produces a table of prediction impacts highlighting dozens of results, storing the models in the database.  

Oracle AI services interpret language and vision

  When it comes to AI services, Oracle offers pre-trained models for a specific-use case, trained and maintained by Oracle, which can be used with your own data – and trained in terms of language and vision. Oracle can identify and translate text from more than 75 languages – ideal for analysing customer service communications – and classifies it into more than 600 categories, as well as recognising and classifying images using labelled data. Hopefully, you will agree with Heli, that there is no need to panic, and you can start experimenting with Machine Learning.