DASCA SDS - Senior Data Scientist

Graphical models are probabilistic models that represent variables and dependencies using graphs:

Markov Random Fields (Option A): Undirected graphical models that capture joint distributions over variables with neighborhood dependencies.

Bayesian Networks (Option B): Directed acyclic graphical models that encode conditional dependencies between random variables.

Geographical Networks (Option C): While they are graphs, they are not probabilistic graphical models used in statistics/ML.

Thus, the correct answer is Option D (Both A and B).

Apache Spark is a distributed data processing engine optimized for big data scenarios. It is specifically designed to handle:

Large-scale datasets spread across clusters.

Massive streaming or batch data pipelines.

Machine learning and graph processing at scale.

Option A: Correct — Spark excels when data is massive and distributed.

Option B: Incorrect — Spark is overkill for small data (Pandas, NumPy, or scikit-learn would be more efficient).

Option C: Incorrect — Spark is not optimized for small datasets.

Option D: Incorrect — since A is valid.

Thus, Spark should be used when data is massive → Option A.

Apache Avro is a widely used framework within the Hadoop ecosystem for data serialization and data exchange.

Option A (Correct): Avro is a compact, fast, binary data serialization format. It allows efficient storage and exchange of structured data.

Option B (Correct): Avro supports Remote Procedure Call (RPC). It provides a framework for RPC communication, making it easier for distributed applications to exchange data across systems.

Option C: Correct, since both statements are true.

Option D: Incorrect because Avro is indeed both a serialization framework and RPC-based.

In data engineering workflows, Avro is valuable because it is schema-based (defined using JSON), highly interoperable, and ensures compatibility across different programming languages. This makes it essential in big data pipelines, Kafka messaging, and Hadoop ecosystem tools.

Thus, the correct answer is Option C (Both A and B).

Data wrangling (also called data munging) refers to transforming raw, messy, or unstructured data into a clean and structured format suitable for analysis.

Option A: Correct. Raw data often contains missing values, duplicates, or irregular formats. Wrangling prepares it for conventional analytics and machine learning.

Option B: Incorrect. Wrangling does not involve "modified meaning"; it focuses on cleaning, structuring, and integrating.

Option C: Incorrect, since only A is correct.

Option D: Incorrect, because wrangling is explicitly described in A.

Thus, the correct answer is Option A.

Structured data refers to highly organized data stored in predefined formats such as rows and columns.

Relational databases (Option A): Structured.

Data tables (Option B): Structured.

CSV files (Option C): Structured because they are tabular with rows and columns.

XML (Option D): Not structured but semi-structured because it stores hierarchical information with tags and attributes.

Thus, the correct answer is Option D (XML).