Dell D-DS-FN-23 Practice Questions with Explanations

Free Dell D-DS-FN-23 practice questions. 23 of them, each with the correct answer, a full explanation, and the reason every other option is wrong.

D-DS-FN-23 practice questions

Question 1

What is a key consideration when preparing a presentation intended for analysts?

  • A. Describe how to implement the model
  • B. Provide talking points to promote or evangelize the project
  • C. Emphasize the business benefits of implementing the model
  • D. Focus on clean simple-to-understand visuals
Show answer and explanation ▾

Correct answer: D

Any presentation of results works best when each visual is clean and makes one point clearly. Simple, uncluttered charts let the audience read the finding quickly instead of working to decode the graphic.

Why the other options are wrong:

  • A. Describing how to implement the model is not the main consideration named for this presentation.
  • B. Talking points that promote the project are aimed at sponsors and stakeholders, not analysts.
  • C. Emphasizing business benefits is the focus of a presentation for executives and the project sponsor.

Question 2

A logistic regression model is built to determine the probability of a credit card borrower defaulting on a credit loan. A threshold value of 0.3 is selected.

Which statement can be used to predict a borrower will default?

  • A. If probability > 0.1, then predict the borrower will default
  • B. If probability < 0.1, then predict the borrower will default
  • C. If probability > 0.3, then predict the borrower will default
  • D. If probability < 0.3, then predict the borrower will default
Show answer and explanation ▾

Correct answer: C

With a threshold of 0.3, the model predicts default when the estimated probability is greater than 0.3, so the prediction rule is probability > 0.3.

Why the other options are wrong:

  • A. 0.1 is not the chosen threshold.
  • B. This uses the wrong threshold and the wrong direction.
  • D. This uses the right threshold but predicts default for low probabilities, which reverses the rule.

Question 3

What are the two data categories that represent qualitative data?

  • A. Ordinal and interval
  • B. Nominal and ordinal
  • C. Ratio and interval
  • D. Nominal and ratio
Show answer and explanation ▾

Correct answer: B

Qualitative (categorical) data is either nominal, with no order, or ordinal, with a ranked order. Interval and ratio data are quantitative.

Why the other options are wrong:

  • A. Interval is a quantitative type.
  • C. Ratio and interval are both quantitative types.
  • D. Ratio is a quantitative type.

Question 4

In hypothesis testing, when does a Type I error occur?

  • A. Null hypothesis is rejected when it is actually false
  • B. Null hypothesis is rejected when it is actually true
  • C. Null hypothesis is accepted when it is actually false
  • D. Null hypothesis is accepted when it is actually true
Show answer and explanation ▾

Correct answer: B

A Type I error is a false positive: the null hypothesis is rejected even though it is actually true.

Why the other options are wrong:

  • A. Rejecting a null hypothesis that is false is a correct decision.
  • C. Accepting a null hypothesis that is false is a Type II error.
  • D. Accepting a null hypothesis that is true is a correct decision.

Question 5

You have been given a task to improve sales force compensation of your organization. As a result of a study, your team decides to classify personnel as follows:

  • Did not meet quota
  • Met quota
  • Exceeded 150% of quota

In which data analytics lifecycle phase should you define these categories for analysis purposes?

  • A. Model building
  • B. Communicate results
  • C. Operationalize
  • D. Model planning
Show answer and explanation ▾

Correct answer: D

Defining the categories and variables to use in the analysis is done in model planning, where you decide which variables and methods the model will use.

Why the other options are wrong:

  • A. Model building runs and fits the model after the categories and variables are chosen.
  • B. Communicate results is the final step of reporting findings.
  • C. Operationalize is about deploying the work, after the model is built.

Question 6

A decision tree is being built. An internal node is being evaluated for partitioning on variables A and B. The entropy of the internal node is 0.8. The entropy for each of the variables is as follows:

  • Variable A: 0.5
  • Variable B: 0.4

Which variable will be used to partition the data and what is the information gain?

  • A. Variable B; information gain is 0.1
  • B. Variable B; information gain is 0.4
  • C. Variable A; information gain is 0.1
  • D. Variable A; information gain is 0.5
Show answer and explanation ▾

Correct answer: B

Information gain is the entropy of the node minus the entropy after the split. Variable A gives 0.8 minus 0.5, or 0.3, and Variable B gives 0.8 minus 0.4, or 0.4, so the tree splits on Variable B with an information gain of 0.4.

Why the other options are wrong:

  • A. The information gain for Variable B is 0.8 minus 0.4, which is 0.4, not 0.1.
  • C. Variable A leaves more entropy after the split, so its gain of 0.3 is lower than that of Variable B.
  • D. Variable A has the lower information gain, and 0.5 is its entropy, not its gain.

Question 7

In association rules, given items X and Y, what does lift measure?

  • A. Percentage of transactions that contain an itemset with X
  • B. Percentage of transactions with X that also contain Y
  • C. Difference in the probability of X and Y appearing together compared with expectations as if they were statistically independent
  • D. How many times more often X and Y occur together than expected if they were statistically independent, expressed as a ratio
Show answer and explanation ▾

Correct answer: D

Lift is the support of X and Y together divided by the product of the support of X and the support of Y. It is a ratio showing how many times more often the items occur together than they would if they were statistically independent.

Why the other options are wrong:

  • A. The percentage of transactions that contain an itemset is support.
  • B. The percentage of transactions with X that also contain Y is confidence.
  • C. A difference between the observed and the expected joint probability is leverage.

Question 8

What are categorized as cluster and workflow management tools for Hadoop?

  • A. Flume, Sqoop, and Storm
  • B. Drill, Hive, and HBase
  • C. Spark, Tez, and Cassandra
  • D. Ambari, Oozie, and Zookeeper
Show answer and explanation ▾

Correct answer: D

Ambari provisions, manages and monitors Hadoop clusters, Oozie schedules and coordinates job workflows, and ZooKeeper provides coordination services for distributed applications. Together they form the cluster and workflow management group of the Hadoop ecosystem.

Why the other options are wrong:

  • A. Flume and Sqoop ingest data and Storm processes streams.
  • B. Drill and Hive are query engines and HBase is a NoSQL data store.
  • C. Spark and Tez are processing engines and Cassandra is a NoSQL database.

Question 9

Consider the following text:

“Stop!” he shouted. “Don’t go there!”

What set of words result from using a tokenizer for punctuation on the text?

  • A. Stop, he, shouted, don, t, go, there
  • B. Stop, he shouted, don, t go there
  • C. Stop, he shouted, dont go there
  • D. Stop, he, shouted, dont, go, there
Show answer and explanation ▾

Correct answer: D

Tokenizing on punctuation strips the quotation marks, exclamation marks, period and apostrophe and splits the text into individual words, giving Stop, he, shouted, dont, go and there.

Why the other options are wrong:

  • A. This result breaks the contraction into two separate tokens, don and t.
  • B. This result leaves several words joined as single tokens.
  • C. This result leaves he shouted and dont go there as multi-word tokens.

Question 10

Which phase of the data analytic lifecycle includes conducting project sponsor interviews and drafting a problem statement?

  • A. Operationalize
  • B. Model planning
  • C. Model building
  • D. Discovery
Show answer and explanation ▾

Correct answer: D

Discovery is the first phase of the data analytics lifecycle. The team learns the business domain, interviews the project sponsor, frames the problem in a problem statement and forms initial hypotheses.

Why the other options are wrong:

  • A. Operationalize delivers final reports, code and a pilot at the end of the project.
  • B. Model planning selects the methods, techniques and variables to use.
  • C. Model building develops training and test datasets and fits the models.

Question 11

In time series analysis, what statement describes a MA(q) process?

  • A. Current deviation from the time series mean depends on the q previous deviations
  • B. Current deviation from the time series mean depends on the quotient q
  • C. Current time series value depends on the q previous values
  • D. Current time series value depends on the fitted polynomial of order q
Show answer and explanation ▾

Correct answer: A

A moving average model of order q expresses the current deviation of the series from its mean as a combination of the q previous deviations, the random shocks from earlier periods.

Why the other options are wrong:

  • B. The value q is the order of the model, a count of lagged terms, not a quotient.
  • C. Dependence on the q previous values of the series itself describes an autoregressive model.
  • D. An MA(q) model does not fit a polynomial of order q to the series values.

Question 12

After running a density plot you realize that the data has a long tail to the right.

What can you do to make the dataset more normally distributed?

  • A. Use a scatter plot to obtain a better picture
  • B. Use a histogram to obtain a better picture
  • C. Apply a square transformation
  • D. Apply a logarithmic transformation
Show answer and explanation ▾

Correct answer: D

A long tail to the right means the data is positively skewed. Taking the logarithm compresses the large values far more than the small ones, which pulls in the tail and makes the distribution closer to normal.

Why the other options are wrong:

  • A. A scatter plot is another way to view data and does not change its distribution.
  • B. A histogram only displays the same skewed distribution.
  • C. Squaring the values stretches large values further and makes the right skew worse.

Question 13

Which R function plots a distribution of a single variable along two different axes?

  • A. table()
  • B. summary()
  • C. density()
  • D. rug()
Show answer and explanation ▾

Correct answer: D

The rug() function adds a rug representation to a plot: a short tick mark for every observation of a single variable, drawn along an axis so the distribution of the values can be seen.

Why the other options are wrong:

  • A. table() counts how often each value occurs and does not plot anything.
  • B. summary() prints descriptive statistics and does not plot anything.
  • C. density() computes a kernel density estimate, which must be passed to plot() to be drawn.

Question 14

What action occurs during feature selection in the model building phase of the data analytics lifecycle?

  • A. Create new combinations of attributes
  • B. Overfit the model to improve prediction accuracy
  • C. Identify the most useful input variables
  • D. Select a superset of variables to shorten training times
Show answer and explanation ▾

Correct answer: C

Feature selection identifies the input variables that contribute most to predicting the outcome and removes the rest. A smaller set of useful variables gives a simpler model that trains faster and is less prone to overfitting.

Why the other options are wrong:

  • A. Creating new combinations of attributes is feature engineering or extraction.
  • B. Overfitting harms accuracy on new data and is something feature selection helps to prevent.
  • D. Feature selection picks a subset of the variables, not a superset.

Question 15

Which Hadoop service responds to requests for compute and memory resources?

  • A. Application Manager
  • B. DataNode
  • C. Scheduler
  • D. Application Master
Show answer and explanation ▾

Correct answer: C

In YARN, the Scheduler is the part of the ResourceManager that allocates cluster resources such as memory and CPU to running applications in response to their resource requests.

Why the other options are wrong:

  • A. This is not the component that allocates compute and memory resources in response to requests.
  • B. A DataNode stores HDFS data blocks and does not allocate compute resources.
  • D. The Application Master requests resources for its application; it does not grant them.

Question 16

What converts SQL-like commands into either Tez, Spark, or MapReduce jobs that are submitted to the Hadoop cluster?

  • A. Pig
  • B. HBase
  • C. Hive
  • D. Mahout
Show answer and explanation ▾

Correct answer: C

Hive provides a SQL-like language, HiveQL, for data stored in Hadoop. It compiles each query into jobs that run on Tez, Spark or MapReduce and submits them to the cluster.

Why the other options are wrong:

  • A. Pig uses its own data flow scripting language, Pig Latin, not SQL-like commands.
  • B. HBase is a NoSQL data store, not a SQL translation layer.
  • D. Mahout is a machine learning library.

Question 17

What is the similarity between the matrix and array data structures in R?

  • A. Both structures can contain only integers
  • B. Both structures can only contain one data type
  • C. Both structures can store multiple data types
  • D. Both structures must be 2-dimensional
Show answer and explanation ▾

Correct answer: B

Matrices and arrays in R are both atomic structures, so every element must be of the same data type. A matrix has two dimensions and an array can have any number of dimensions.

Why the other options are wrong:

  • A. Both structures can hold other types, such as character or logical values, as long as all elements share one type.
  • C. Neither structure can mix data types; lists and data frames are used for that.
  • D. Only a matrix is limited to two dimensions; an array can have more.

Question 18

Which visualization technique should be avoided?

  • A. Using a small number of contrasting colors to draw distinctions
  • B. Using tables of numbers to present all of the data visually
  • C. Achieving a high data-ink ratio
  • D. Using visuals to illustrate key points
Show answer and explanation ▾

Correct answer: B

A large table of numbers forces the audience to find the pattern themselves. Good visualization selects the data that supports the point and shows it in a chart, instead of displaying all of the data as raw figures.

Why the other options are wrong:

  • A. A small number of contrasting colors is a recommended way to draw distinctions.
  • C. A high data-ink ratio is a goal of good chart design.
  • D. Using visuals to illustrate the main points is recommended practice.

Question 19

Which analytic technique would be appropriate to estimate home sale price in U.S. dollars as a function of square footage, number of bedrooms, and lot size?

  • A. Time series analysis
  • B. Linear regression
  • C. Naïve Bayesian classification
  • D. K-means clustering
Show answer and explanation ▾

Correct answer: B

Sale price in dollars is a continuous outcome, and square footage, bedrooms and lot size are numeric inputs. Linear regression models a continuous outcome as a linear function of input variables like these.

Why the other options are wrong:

  • A. Time series analysis models values observed over time, not a price as a function of property attributes.
  • C. Naive Bayesian classification assigns categorical labels, not a dollar amount.
  • D. K-means clustering groups similar records and does not predict an outcome value.

Question 20

How should project results be communicated to executives and the project sponsor?

  • A. Focus on business outcomes and benefits
  • B. Demonstrate your technical prowess to establish credibility
  • C. Provide model performance visualizations
  • D. Emphasize coding details and technical requirements
Show answer and explanation ▾

Correct answer: A

Executives and the project sponsor want to know what the project means for the business. The presentation should lead with the business outcomes and benefits and leave out the technical detail.

Why the other options are wrong:

  • B. Displaying technical skill does not tell executives what the project delivers.
  • C. Model performance visualizations belong in the presentation for analysts and data scientists.
  • D. Coding details and technical requirements are for the technical team.

Question 21

What is a business driver for Big Data analytics adoption?

  • A. Implement the latest technology and tools
  • B. Maintain existing data silos
  • C. Identify new business opportunities
  • D. Ensure the analysts work in isolation
Show answer and explanation ▾

Correct answer: C

Organizations adopt Big Data analytics to solve business problems, and identifying new business opportunities is one of those drivers, along with optimizing operations, managing risk and meeting compliance needs.

Why the other options are wrong:

  • A. Adopting new technology for its own sake is not a business driver.
  • B. Data silos are an obstacle that Big Data analytics aims to remove.
  • D. Analysts working in isolation is a weakness of traditional setups, not a goal.

Question 22

What characterizes the Hadoop Distributed File System?

  • A. Peer to peer system designed to run on custom designed hardware
  • B. Peer to peer system designed to run on commodity hardware
  • C. Master / slave system designed to run on custom designed hardware
  • D. Master / slave system designed to run on commodity hardware
Show answer and explanation ▾

Correct answer: D

HDFS uses a master and slave design: a NameNode holds the file system metadata and DataNodes store the data blocks. It is built to run on clusters of low-cost commodity hardware and to tolerate hardware failure through replication.

Why the other options are wrong:

  • A. HDFS is not a peer to peer system and does not need custom hardware.
  • B. HDFS has a NameNode master and is not a peer to peer system.
  • C. HDFS is designed for commodity hardware, not custom designed hardware.

Question 23

After which phase of the data analytics lifecycle should you determine if the model needs any recalibration?

  • A. Model planning
  • B. Data preparation
  • C. Discovery
  • D. Operationalize
Show answer and explanation ▾

Correct answer: D

In the Operationalize phase the model is run in a pilot on production data. Its performance there shows whether the model needs to be adjusted or retrained before and after full deployment.

Why the other options are wrong:

  • A. Model planning takes place before any model has been built or run.
  • B. Data preparation conditions the data and comes before modeling.
  • C. Discovery frames the business problem at the very start of the project.

That was 23 of 59.

The full Dell D-DS-FN-23 pack has all 59 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. $39.00 USD, paid once, with free monthly updates and a pass-or-your-money-back guarantee.

Get the full pack

Not ready yet? Get the first 23 free as one PDF by email. Get the Half Pack →

About the D-DS-FN-23 exam

These 23 Dell D-DS-FN-23 questions are free here, and as a free PDF by email.

They are drawn from the same bank as the full D-DS-FN-23 pack, which has 59 questions in total.

Back to blog