Exploratory Data Analysis

Question 1

What are the 3 major types of data?

Answer 1

-Numerical
-Categorical
-Ordinal

Question 2

In Timeseries data, what is Trend and Sazonality?

Answer 2

A Trend is the overall expected change of datapoints based on time passed, while seasonality is a repeating behavior that cycles as time passes

Question 3

What are the data formats supported by Amazon Athena?

Answer 3

-CSV
-JSON
-Parquet
-Avro
-ORC

Question 4

How does Athena’s billing work?

Answer 4

-You pay for each 5TB of data scanned

Question 5

True or False: Data transfers between S3 and Athena are unencrypted

Answer 5

False, they are encrypted using TLS

Question 6

What are Amazon Quicksight’s accepted data origins?

Answer 6

-Excel, CSV, TSV and log files (Local or S3)
-Athena
-Redshift
-RDS / Aurora
-EC2 hosted databases
-AWS IoT Analytics

Question 7

What is SPICE in a Quicksight context?

Answer 7

It is an in-memory, high performance calculation engine used by quicksight to perform it’s queries.

Question 8

How much memory does each Quicksight user have access to on SPICE?

Answer 8

10GB

Question 9

What are the machine learning insights available on Quicksight?

Answer 9

-Anomaly detection
-Forecasting
-Auto-narrative

Question 10

What is Quicksight Q?

Answer 10

It is a GenAI functionality on Quicksight that can analise your data and generate insights to questions asked to it

Question 11

What is a Quicksight paginated report?

Answer 11

It is a report that is created in a format designed to be printed

Question 12

True or False: Quicksight was not designed for perfoming ETL, Glue being recommended for that instead

Answer 12

True

Question 13

True or False: Quicksight supports user creation via IAM or email, but not 2FA

Answer 13

False, it does support 2FA

Question 14

Ture or False: You can restrict Athena Access to S3 data through the usage of bucket policies

Answer 14

True

Question 15

What is EMR?

Answer 15

It is an AWS service that allows you to run a Managed Hadoop framework on EC2 instances

Question 16

True or false: EMR includes Spark, Hive, Flink, Presto and HBase

Answer 16

True

Question 17

In the architecture of an EMR cluster, what are the existing node types?

Answer 17

In an EMR cluster, there are the following nodes:
-Master node: Manages the cluster
-Core node: Hosts HDFS data and tuns tasks
-Task nodes: Runs task while not storing any data (Good use of spot instances)

Question 18

True or False: EMR Serverless lets AWS scale your nodes automatically

Answer 18

True

Question 19

What are the types of EMR data storage?

Answer 19

-HDFS
-EMRFS: Access S3 as if it was HDFS
-Local File System
-EBS for HDFS

Question 20

What is the EMRFS Consistent View setting for EMR data storage?

Answer 20

Optional feature used to mantin consistency on S3 data using DynamoDB

Question 21

What models are available on SparkMLLib?

Answer 21

-Classification: Logistic Regression and Naive Bayes
-Regression
-Decision trees
-Recommendation engine (ALS)
-Clustering (k-means)
-LDA (topic modelling)
-SVD, PCA, statistics
-ML workflow utilities (pipelines, fetaure transformation, persistence)

Question 22

What is the recommended size for EMR mater nodes?

Answer 22

m5.large if <50 nodes, m5.xlarge otherwise

Question 23

What does HDFS mean?

Answer 23

Hadoop Distributed File System

Question 24

What is the difference between a Transient and a Long-running EMR cluster?

Answer 24

A transient cluster is automatically terminated after its tasks are run, while a Long-running one is terminated manually by the user

Question 25

True or False: when you terminate an EMR Cluster using HDFS storage, the files are backed up to EBS

Answer 25

False, all the HDFS file is ephemeral and lost

Question 26

What are better ways than mean of imputing missing data?

Answer 26

-KNN mean (mean o N closest entries) -Deep Learning (Good for categorical data) -Regression (Advanced technique: MICE)

Question 27

What is SMOTE?

Answer 27

It is an oversampling technique where new artificial entries are generated to solve biased datasets. Also Undersamples majority class.

Question 28

What is a possible statistical algorithm that can be used to detect outliers?

Answer 28

Random Cut Forest

Question 29

What is binning?

Answer 29

Grouping multiple entries with similar values together.

Question 30

What is quantile binning?

Answer 30

It is a strategy where you perform bining based on the value distribution, making sure all bins have the same size.

Question 31

What is Sagemaker Ground Truth?

Answer 31

It is an AWS service used for labeling data for ML

Question 32

What is the difference between Sagemaker Ground Truth and Mechanical Turk?

Answer 32

Ground truth is directed specifically to data labelling, while Mechanical Turk focuses in any task that has to be manually performed by humans.

Question 33

Ture or False: Sagemaker Ground Truth works 100% on human labour.

Answer 33

False, as images are labelled, ground truth crates it's own model, only requesting human assistance for instances where it is not sure of it's own classifications.W

Question 34

What is Sagemaker Ground Truth Plus?

Answer 34

It is a feature of Ground Truth where the image workflow and labelling is managed by AWS experts and then delivered to the customer.

Question 35

What are other AWS services that can be used to generate training labels?

Answer 35

AWS Rekognition (image) and AWS Comprehend (text)