Practice Exam: Oracle Machine Learning using Oracle Autonomous

Database

1. When managing models using the OML Models interface on Autonomous

Database, what are the three operations a user can perform on models?

Change the owner of a previously deployed in-database model.

Undeploy a previously deployed in-database model.

Change the deployment date of a previously deployed in-database model.

Deploy an existing in-database model as a REST endpoint in OML Services.

Delete an existing in-database model.

2. Change the namespace of a previously deployed in-database model.

You create an in-database Decision Tree model using OML4R, and you want
to review any of the leaf nodes in the tree that have more than 50 records.
How can you achieve this using OML4R?

dt_mod <- ore.odmDT(target ~ ., INPUT_DATA, nodes = 50)

summary(dt_mod)

dt_mod <- ore.odmDT(target ~ ., INPUT_DATA)

dt_mod$nodes[dt_mod$nodes$row.count > 50,]

dt_mod <- ore.odmDT(target ~ ., INPUT_DATA)

predict(dt_mod$nodes$row.count > 50)

dt_mod <- ore.odmDT(target ~ ., INPUT_DATA, nodes = 50)

summary(dt_mod$nodes)

3. You need to compute the correlation of two numeric columns X, Y in a

database table referenced by a data frame proxy object DF. The correlation
here is defined as E[(X – mean(X))*(Y – mean(Y))]/cov(X,Y).
What are two ways to do it in OML4R and OML4Py?
a) In OML4Py, DF['X', 'Y'].corr(method = 'pearson').
b) In OML4R, ore.corr(DF,var='X,Y', stats = 'spearman').
c) In OML4R, ore.corr(DF,var='X,Y', stats = 'pearson').
d) In OML4Py, DF['X', 'Y'].corr(method = 'spearman').
4. Which two statements are FALSE regarding the OML Services REST API
access token?

The token is tied to the OML user who authorizes with the OML user
credential.

The token is valid for 1 hour.

An expired token can be refreshed.

Each token can be used many times.

A token can be refreshed up to 8 hours.

A revoked token can be refreshed.

A token can be revoked.

5. What is the primary function of OML Services?

To provide cloud storage solutions

To manage user access and permissions

To deploy and manage machine learning models using REST endpoints

To offer web hosting services

6. You have produced several regression models, and now you are asked to
select the best model of the set. Which two types of statistics would you use
to evaluate regression models?

Root mean squared error

Mean absolute error

True positive rate

Confusion matrix statistics

7. Which four classification algorithms are supported in Oracle Data Miner?

Decision Tree

K-Nearest Neighbors

Generalized Linear Model

Factor Analysis

NaïBayes

Support Vector Machine

 Neural Networks

8. Which two Oracle Data Miner nodes focus on data exploration and
visualization?

SQL Query node

Explore Data node

Analyze node

Graph node

Class Build node

Data Visualization node

9. For which reason does a data scientist perform "feature engineering" when
building models?

Data scientists need clean data and prepared data. Feature engineering,
required for the algorithms, removes missing values and normalizes data so
the features are between -1 to 1 or 0 to 1 ranges.

Machine learning algorithms require a reduced and prioritized collection of

input features. Oracle Machine Learning uses the Minimum Description
Length (MDL) algorithm for feature selection (also known as attribute
importance).

Machine learning algorithms require a reduced and prioritized collection of

input attributes. Oracle Machine Learning uses the Non-Negative Matrix
Factorization (NMF) algorithm for feature extraction.

Feature engineering derives new features from existing features to build

better models. For example to target High_Life_Time_Value Customers,
rather than using Purchase_Amount, you could create a new attribute:
"Count_When_Purchase_Exceeds_$500_in_12_month_period."

10.Given a table mining_data_test with columns CUST_ID, YRS_RESIDENCE,

CUST_GENDER, and AFFINITY_CARD. You built a classification model using
YRS_RESIDENCE and CUST_OWN_OR_RENT to predict AFFINITY_CARD.
Which three code sequences compute the confusion matrix after the model
is built?

In OML4SQL, the model is built with the name nb_sh_clas_sample

CREATE TABLE nb_apply_results AS
SELECT cust_id,
PREDICTION(nb_sh_clas_sample USING *) prediction,
PREDICTION_PROBABILITY(nb_sh_clas_sample USING *) probability
FROM mining_data_test;
 BEGIN
DBMS_DATA_MINING.COMPUTE_CONFUSION_MATRIX (
accuracy => v_accuracy,
apply_result_table_name => 'nb_apply_results',
target_table_name => 'mining_data_test',
case_id_column_name => 'cust_id',
target_column_name => 'affinity_card',
confusion_matrix_table_name => 'nb_confusion_matrix',
score_column_name => 'PREDICTION',
score_criterion_column_name => 'PROBABILITY'
score_criterion_type => 'PROBABILITY');
END;

SELECT * from nb_confusion_matrix;

In OML4Py, the model is built as dt_mod

RES_DF = dt_mod.predict(mining_data_test [['YRS_RESIDENCE',
'CUST_OWN_OR_RENT']],
supplemental_cols = mining_data_test ['AFFINITY_CARD'])
RES_DF.crosstab('AFFINITY_CARD', 'PREDICTION')

In OML4R, the model is built as dt_mod

mining_data_test$PRED <- ore.predict(dt_mod, mining_data_test)
table(mining_data_test$PRED, mining_data_test$AFFINITY_CARD)

In OML4Py, the model is built as dt_mod

RES_DF = dt_mod.predict(mining_data_test [['YRS_RESIDENCE',
'CUST_OWN_OR_RENT']],
supplemental_cols = mining_data_test ['AFFINITY_CARD'],
proba = TRUE)
RES_DF.crosstab('AFFINITY_CARD', 'PROBABILITY')

11.On Autonomous Database, you are using OML Notebooks and want to
schedule a recurring notebook execution. The lead data scientist wants the
script run five times every other day starting today at 11:00 PM, but only
retry twice if it should fail, and pad the timeout to account for possible
longer execution times. Normally, the execution takes 20 minutes.

Repeat Frequency: 2 days

Maximum Number of Runs: 5
Maximum Failures Allowed: 2
Timeout in minutes: 40 minutes
 Repeat Frequency: 5 days
Maximum Number of Runs: 2
Maximum Failures Allowed: 2
Timeout in minutes: 20 minutes

Repeat Frequency: 2 days

Maximum Number of Runs: 5
Maximum Failures Allowed: 2
Timeout in minutes: 10 minutes

Repeat Frequency: 2 weeks

Maximum Number of Runs: 2
Maximum Failures Allowed: 5
Timeout in minutes: 60 minutes

12.A customer's data science team consists of both R users with on-premises
development and cloud production environments. Data is periodically
synchronized between the Oracle Base Database Service and on-premises
Oracle Database instances.
Which scenario do you recommend to support their project for performance,
scalability, and maintainability while maintaining data security?

Use OML4R with Oracle Database on premises to develop and test R user-
defined functions, then test and deploy these scripts in Oracle Base
Database Service. Use the SQL API to invoke the user-defined R functions
from applications.

Use open source R with ROracle to extract database data and develop
scripts to meet project goals. Manually spawn R engines from the
application to run these scripts in production.

Use OML4R with Oracle Database on premises to develop and test user-
defined R functions, then use open source R with ROracle to connect to
Oracle Base Database Service for testing and deploying these scripts.

Use open source R with ROracle to extract database data and develop
scripts to meet project goals. Then, store these same scripts in the Oracle
Base Database Service R script repository. In R, load these scripts into the R
engine for execution.

13.Which four parameters are supported/valid to use in oml.connect(……)

provided by OML4Py to establish an OML4Py connection to an Oracle
Database instance?

USER

PASSWORD

DSN
 AUTOML

RESTRICTION

WALLET

14.Where are notebooks managed by OML Notebooks stored for a given user?

Oracle Cloud Infrastructure Object storage

Oracle Cloud Infrastructure File storage

Oracle Cloud Infrastructure Block storage

Oracle Cloud Infrastructure Archive storage

Autonomous Database user's own schema

15.Which statement about automl.FeatureSelection in OML4Py is correct

syntax?

from oml.automl import FeatureSelection

fs = FeatureSelection(mining_function = 'classification', score_metric =
'accuracy')

from oml.automl import FeatureSelection

fs = FeatureSelection(mining_function = 'classification', metric = 'accuracy')

from oml.automl import FeatureSelection

fs = FeatureSelection(mining_function = 'clustering', score_metric =
'accuracy')

from oml.automl import FeatureSelection

fs = FeatureSelection(mining_function = 'clustering', metric = 'accuracy')

16.Which is a FALSE statement regarding Oracle Machine Learning (OML)?

OML provides univariate and multivariate statistics.

OML provides scalable statistical functions through OML4Py and OML4R.

OML provides integration with open source Python and R statistical analysis
functions.

OML needs a separate data visualization tool for creating visualizations.

17.Which OML component on Autonomous Database provides a REST API?

OML4Py

OML4SQL

Oracle Data Miner

Oracle Data Pump

18.In the OML Model Monitoring UI, which two of the following are not a
prediction statistic provided by the monitor?

Population Stability Index

Standard Deviation

Variance

Mean

Median

19.What are two advantages of Oracle Machine Learning for Python?

Operates on database data from Python without using SQL

Maximizes data movement

Supports the use of additional Python packages to complement in-database

functionality through embedded Python execution

Cannot automate common Machine Learning tasks

20.What is Machine Learning?

A utility for unloading data and metadata into a set of operating system files
called a dump file set

A field of AI that involves developing algorithms and statistical models that

enable learning from and making predictions or decisions using data,
without being explicitly programmed.

A comprehensive set of services that create, maintain, manage, and

monitor one or more standby databases to enable production Oracle
databases to survive disasters and data corruptions.

21.What are two key features of Machine Learning?

Prevents data loss and down time

Analyzes small volumes of data

 Supports a range of techniques including classificatoin, regression, and
clustering

Enables discovery of patterns without explicit programming

22.Which three types of permissions can a user be granted to a workspace?

Developer

Viewer

Contributor

Member

Downloader

Manager

23.What is a notebook?

It lists all the jobs created, along with the job name, owner of the job, last
start date, next run date, status, and schedule.

It is a collaborative interface where you can wite SQL, R, and Python code
and document your work using markdown.

It contains all the elements associated with a project.

It is the space within OML where your projects reside.

24.Which two can you access from the Oracle Machine Learning home page?

Your Autonomous Database instance

Quick links to important interfaces

A log of your recent high-level activities

Database actions

25.Which two components on Autonomous Database Serverless can be used to

develop and deploy SQL, R, and Python code?

Workspaces

Projects

OML Notebooks

SQL Developer Web

26.Which four languages are supported by Oracle Machine Learning?

 SQL

Python

Swift

PL/SQL

27.Oracle Autonomous Database is designed for the administration of a

database with SQL commands executed by using Machine Learning.

TRUE

FALSE

28.Which two can you access from the Oracle Machine Learning home page?

A log of your recent high-level activities

Your Autonomous Database instance

Database actions

Quick links to important interfaces

29.Which step in the Oracle AutoML pipeline aims to find the smallest sample
size that adequately represents the full data set?

Model Tuning

Adaptive Sampling

Algorithm Ranking

Feature Selection

30.Which step in the Oracle AutoML pipeline determines the settings or

hyperparameters that produce a better quality model within the constraints,
such as time, specified?

Algorithm Ranking

Adaptive Sampling

Feature Selection

Model Tuning
31.Which class would you use to automatically select the best Oracle Machine
Learning algorithm and tune the model?

oml.automl.ModelSelection

oml.automl.ModelTuning

oml.automl.FeatureSelection

oml.automl.AlgorithmSelection

32.You must bind a notebook to an interpreter to fetch data from a database or

any data source. Which is the default interpreter?

Low

Medium

High

Extreme

33.Which statement is false?

You can create a backup of a notebook.

You can delete the selected version of your notebook.

You can create a new notebook from the selected notebook version.

You cannot restore your notebook to the older version.

34.For which of the following is OML Notebooks NOT a web-based interface?

Data analysis

Data discovery

Data insights

Data visualization

Collaboration

1.

Which is NOT one of the three types of templates?

Personal templates

Shared templates

Example templates
Fixed templates

2.

Which task cannot be performed with personal templates?

Run all paragraphs in the template notebook

Rename the template notebook

Create a notebook from the available templates

Delete selected notebook templates.

Save selected notebook as a shared template

3.

Which is NOT an action that results in the Shared Templates page tracking
notebook templates?

Liking templates

Deleting templates

Creating notebooks from templates

Viewing templates

1.

Which statement is NOT true about jobs?

On the Jobs page, you can create, duplicate, stop, and delete jobs.

You can edit an existing job.

You can version jobs.

Jobs enable you to create schedules to run notebooks.

2.

What is the maximum character limit for a job name?

64 bytes

100 bytes

128 bytes

130 bytes

3.

Which is NOT viewable on the Jobs page?

Jobs log

All the jobs created

The next run date

Job status

1.

Which statement about model monitoring is NOT true?

Model monitoring helps detect deteriorating model accuracy and provide

insights into underlying causes

Model monitoring uses model accuracy trends and model feature impact to
guide model rebuilding

Model monitoring automatically addresses data quality issues affecting

model accuracy

2.

When creating a model monitor through the UI, you can specify data
monitoring as well.

TRUE

FALSE

3.

Data monitoring is only useful in the contect of machine learning.

TRUE

FALSE

1.

Which task can be performed by a developer?

Create a user account and password.

Create and schedule jobs.

Create connection groups to allow connectivity to the database.

Remove all users.

2.

Which statement about the Compute Resources page is NOT true?

Compute resource refers to the services to which a notebook Interpreter
connects.

The Compute Resources page displays the list of compute resources.

The Compute Resources page displays the name of each resource and last-
updated details.

You do not need the Administrator role to access the Compute Resources
page.

3.

An administrator can create, edit, and remove user accounts in OML.

TRUE

FALSE

1.

Which statement about the Compute Resources page is NOT true?

Compute resource refers to the services to which a notebook Interpreter

connects.

The Compute Resources page displays the list of compute resources.

The Compute Resources page displays the name of each resource and last-
updated details.

You do not need the Administrator role to access the Compute Resources
page.

2.

Which task can be performed by a developer?

Create a user account and password.

Create and schedule jobs.

Create connection groups to allow connectivity to the database.

Remove all users.