Just like with graph algorithms, Oracle 23c has more than 30 algorithms ready to use for machine learning models.
Did anyone ask for AI? Machine learning is AI, and of course, there is more AI beyond just machine learning, but these are the interesting pieceswe can perform in the Oracle Database.

We even discussed that the Oracle Database statistic collection is now leveraging machine learning algorithms to make real-time statistics more efficient and predict information about data changes and growth.
Machine learning is there to answer additional questions and has the ability to analyze large
volumes of data.
What questions are being asked of your data? Have you been able to answer the business needs with reports and queries? If not, it is time to investigate implementing graph and machine learning and using the data for business value.

Machine learning can automatically discover patterns and create actionable information. Using the PL/SQL package DBMS_DATA_MINING, machine learning models can be built quickly.
You will need to gain some understanding of the types of algorithms such as classification, which can be used for predictions about customers.

In Autonomous Database, Oracle AutoML is available and can help you select the right types of algorithms. AutoML helps nonexpert users leverage machine learning in the Oracle Database.

Here is the process for AutoML:
• Auto algorithm selection: Finds the best algorithm from in-database algorithms
• Adaptive sampling: Identifies and adjusts the right sample size
• Auto feature selection: Improves accuracy and performance
• Auto model tuning: Improves models with automated tuning

AutoML provides a quick way to get started with machine learning and test the built-in algorithms and build models without completely understanding all of the science. However, with the different skill sets, the packages are available.
For example, let’s take our customer list, and based on their begin date as a customer, figure out if they are at risk of not purchasing additional items or leaving as a customer:

SQL> create or replace view vw_customer_longterm asselect c.customer_id, c.customer_name, decode(to_char(customer_begin_ date,’YYYY’,’2023′,1,0) cust_date_value,decode(nvl(o.order_id,0),0,0,1) order_value from customer c left outer join orders o on o.customer_id=c.customer_id; View created.

SQL> declarev_setlst dbms_data_mining.setting_list; beginv_setlst(dbms_data_mining.algo_name) := dbms_data_mining.algo_support_ vector_machines;v_setlst(dbms_data_mining.prep_auto) :=dbms_data_mining.prep_auto_on; dbms_data_mining.create_model2(model_name =>’FIND_LONGTERM’, mining_function => ‘CLASSIFICATION’,data_query => ‘select * frommmtest.customer_longterm’, set_list => v_setlst,case_id_column_name => ‘CUSTOMER_ID’, target_column_name => ‘CUST_DATE_VALUE’); end;/\

SQL> create table customer_longterm as select customer_id,prediction(FIND_LONGTERM usering*) cust_likely_to_leave, prediction_details(FIND_LONGTERM using *) PDfrom vw_customer_longterm;

Machine learning can confirm observations and find new patterns. Discovery and predictive relationships are ways to start harnessing the data by pulling information out of the data.