SDS Sample Questions Answers

A Bernoulli random variable is the simplest form of discrete random variable.

It can take only two values:

1 with probability p (success).

0 with probability (1 – p) (failure).

Since the outcomes are finite (binary), it is a discrete random variable.

Option B (Continuous): Incorrect, as continuous variables take values from an interval (e.g., real numbers).

Option C and D: Incorrect, as Bernoulli is always discrete, never continuous.

Thus, the correct answer is Option A (Discrete random variable).

Exploratory analytics (often referred to as Exploratory Data Analysis – EDA) is a fundamental step in data science, enabling practitioners to discover initial insights, detect anomalies, and understand the structure of datasets before applying predictive or prescriptive modeling.

Option A (Understand the data content): Correct. EDA techniques (descriptive statistics, summary tables, profiling) reveal missing values, data types, and distributions.

Option B (Gain a high-level understanding of relationships): Correct. Correlation analysis, scatter plots, and cross-tabulations help identify dependencies between variables.

Option C (Understand patterns in the data): Correct. Visualization and clustering methods help discover hidden structures, seasonalities, and outliers.

Since exploratory algorithms contribute to all of these objectives, the correct answer is Option E (All of the above).

HDFS (Hadoop Distributed File System) provides quota management to control and monitor resource usage across directories:

Name Quotas (Option A): Limits the number of files and directories that can be created in a given HDFS directory. Helps prevent excessive metadata growth.

Space Quotas (Option B): Limits the total disk space consumed by files within a directory. Helps in capacity planning and avoiding storage overuse.

Since HDFS supports both types, the correct answer is Option C (Both A and B).

Apache Spark supports multiple programming languages for developing distributed applications:

Java (Option A): Supported through Spark’s JVM-based APIs.

Scala (Option B): Spark is natively written in Scala, and Scala APIs provide full functionality.

Python (Option C): Supported via PySpark, enabling Python developers to leverage Spark.

Additionally, Spark also supports R and SQL-like queries, making it versatile for data scientists and engineers.

Thus, the correct answer is Option D (All of the above).

Unsupervised learning is a machine learning approach where no labeled outputs are provided. The algorithm discovers patterns or structures directly from raw data.

Option A (Detecting regularities): Correct. Unsupervised learning identifies hidden structures such as clusters, associations, and dimensionality reductions (e.g., k-means clustering, PCA).

Option B (Detecting irregularities): Correct. Outlier detection is also a part of unsupervised learning, often used in anomaly detection (e.g., fraud detection, intrusion detection).

Option C: Correct, since unsupervised learning helps detect both regularities (clusters, groups) and irregularities (outliers, anomalies).

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

LaTeX is a high-quality typesetting system widely used in academia, particularly in scientific publishing.

Statement i: Correct. LaTeX is widely used to prepare manuscripts for scientific journals, theses, and technical reports.

Statement ii: Correct. LaTeX is a markup language (similar to HTML in concept) that compiles into formatted PDFs/documents.

Statement iii: Correct. LaTeX is a standard for publishing scientific papers due to its ability to handle complex mathematical equations, references, and formatting.

Thus, all three statements are true → Option B (i, ii, iii).

Machine Learning has broad applications across industries and technologies:

Fraud Detection (Option A): Detecting anomalies in financial transactions, credit card usage, and cybersecurity threats.

Web Search Results (Option B): Ranking algorithms (e.g., Google’s PageRank enhanced by ML techniques) improve relevance of search queries.

Real-time Ads (Option C): Online ad systems use reinforcement learning and recommendation models to target ads dynamically.

Pattern & Image Recognition (Option D): ML (especially deep learning) powers facial recognition, handwriting recognition, medical imaging, etc.

Since ML is used in all these applications, the correct answer is Option E (All of the above).

JSON (JavaScript Object Notation) is a lightweight data-interchange format widely used for storing and exchanging structured or semi-structured data. JSON allows hierarchical (tree-like) structures, such as nested objects and arrays, to be serialized into a textual representation.

Option A (Plain text format): Correct. JSON files are stored as plain text, making them human-readable and language-independent.

Option B (Plain string format): Correct. JSON objects are transmitted as strings across networks (e.g., via APIs, RESTful services).

Option C: Incorrect. JSON does not serialize into “any format,” but specifically into text/string-based formats.

Option D: Correct. Since JSON is both plain text and transmitted as string format, the right answer is both A and B.

Option E: Incorrect.

Thus, JSON serializes hierarchical data into plain text and string formats.

A daily stand-up meeting (also called a daily Scrum) is a short meeting (usually 15 minutes) that Agile teams hold to synchronize progress. Its agenda is structured around three key questions:

What was accomplished yesterday? (Progress review).

What is planned for today? (Work alignment).

What impediments or roadblocks exist? (Barriers identification).

This process enhances transparency, communication, and accountability, ensuring the team can quickly address obstacles and stay aligned with sprint goals.

Option A: Correct — yesterday’s work is discussed.

Option B: Correct — today’s planned tasks are outlined.

Option C: Correct — roadblocks are highlighted.

Option D: Incomplete since it misses C.

Option E: Correct — covers all agenda items.

Thus, the correct answer is Option E (All of the above).

Grid computing is a distributed computing model where resources (CPU, memory, storage) are pooled across multiple systems to solve large-scale problems.

Option A (Divide): Middleware helps allocate or divide resources dynamically to different tasks.

Option B (Combine): Middleware integrates diverse resources into a unified system, making them accessible for parallel computing.

Option C: Correct — middleware is the “glue” that enables both combining and dividing resources seamlessly across distributed nodes.

Option D: Incorrect.

Thus, grid computing middleware both combines and divides resources, making Option C correct.

Graphical models represent relationships between objects using nodes (entities) and edges (relationships). Examples include:

Road maps (Option A): Nodes = intersections, Edges = roads.

Electrical circuits (Option B): Nodes = components, Edges = connections.

Computer networks (Option C): Nodes = devices, Edges = connections.

Geographical networks (Option D): Nodes = locations, Edges = transport or connectivity.

However:

Flow charts (Option E): These represent process flows, not structural networks of entities and relationships. They are procedural diagrams, not graphical models in the statistical/graph-theory sense.

Thus, the correct answer is Option E (Flow charts).

Evolutionary algorithms are optimization techniques inspired by natural selection. They are often applied in neural network optimization for evolving architectures, weights, or hyperparameters.

Genetic Algorithm (Option A): Uses selection, crossover, and mutation to evolve neural network parameters.

Genetic Programming (Option B): Evolves computer programs or expressions; can be applied to evolve neural network structures.

Evolution Strategy (Option C): Focuses on optimization using mutation and adaptation of parameters.

Since all are used as evolutionary approaches in neural networks, the correct answer is Option D (All of the above).

The Chief Data Monetization Officer (CDMO) is a C-level executive role emerging in data-driven organizations. Their focus is on turning data into business value and revenue growth.

Statement i: Correct. The CDMO is responsible for value realization from data and analytics investments.

Statement ii: Correct. The CDMO owns decisions around investment prioritization for data initiatives.

Statement iii: Correct. Since monetization is directly linked to business performance, the CDMO must be accountable for revenue and margin impact.

Hence, all three statements are correct, making the right answer Option C (All of the above).

Python’s built-in data containers are widely used in data science and programming:

Lists (Option A): Ordered, mutable sequences.

Tuples (Option B): Ordered, immutable sequences.

Dict (Option D): Dictionaries, key-value data structures implemented using hash tables.

LinkedList (Option C): Python does not provide a native linked list as a built-in container. While linked lists can be implemented manually or through external libraries (e.g., collections.deque for efficient appends/pops), they are not part of Python’s main native containers.

Thus, the correct answer is Option C (LinkedList).

TAR (Tape Archive) is a widely used archive file format in Unix/Linux environments. It is used to combine multiple files into a single archive file (with extension .tar).

Option A: Correct. TAR is specifically designed for archiving.

Option B (CSV): Incorrect. CSV (Comma-Separated Values) is a tabular text data format.

Option C (ARV): Incorrect — no such format.

Option D (Text): Incorrect. Though TAR may contain text files, the TAR format itself is not plain text but an archive format.

Option E: Incorrect since Option A is valid.

Thus, TAR is an Archive file format.

The Big Data Vision Workshop is an early-phase framework designed to help organizations shape their data-driven transformation journey. It is particularly beneficial when:

Option A: Organizations want to leverage big data but lack clarity on where to start.

Option B: Organizations already have large volumes of data but struggle to derive monetization strategies from it.

Option C: Organizations want to identify use cases where data and analytics can enhance or even redefine their business models.

Since all three statements apply, the correct answer is Option E (All of the above).

In hypothesis testing, two main types of errors are defined:

Type I Error (Option A): Occurs when the null hypothesis (H₀) is true, but we incorrectly reject it. This is known as a false positive. Example: Concluding a drug is effective when it is not.

Type II Error (Option B): Occurs when the null hypothesis (H₀) is false, but we fail to reject it. This is a false negative. Example: Concluding a drug has no effect when it actually does.

Logical Error / Hypothesis Error (Options C and D): Not standard terms in statistical hypothesis testing.

Thus, the “wrong negation of a true null hypothesis” refers to a Type I Error (false positive).

Regression is a supervised learning technique where a model estimates the relationship between input features (independent variables) and an output (dependent variable).

Option A: Correct. The learning process involves optimizing model parameters (e.g., coefficients in linear regression) to minimize approximation error. Common loss functions include Mean Squared Error (MSE) or Mean Absolute Error (MAE).

Option B: Correct. Minimizing error enables the model to produce the closest possible outcomes to the actual observed values, ensuring accurate predictions.

Option C: Correct, since both A and B are true.

Option D: Incorrect.

Thus, regression optimization in machine learning aims to minimize approximation error and generate closest possible outcomes, making Option C the correct answer.

Use Case Development in data science projects involves identifying business needs and mapping analytics to business decisions. The standard steps include:

Option A: Understanding the key initiatives or challenges.

Option B: Identifying the key stakeholders.

Option C: Capturing the decisions stakeholders must make.

Option E: Brainstorming the questions stakeholders need answered to support decisions.

However:

Option D (Brainstorm the outcomes stakeholders need to answer): Incorrect phrasing. It is not “outcomes” that are brainstormed but questions and decisions.

Thus, the correct answer is Option D.

Self-driving cars (autonomous vehicles) are an application of Reinforcement Learning (RL) in machine learning:

In RL, an agent (car) interacts with an environment (roads, obstacles, traffic) and learns to maximize rewards (e.g., safe driving, efficient navigation).

The system improves performance through trial-and-error learning, guided by reward signals such as staying in a lane or avoiding collisions.

Supervised learning (A): Used in some supporting tasks like image recognition (e.g., identifying stop signs), but not the core paradigm for self-driving.

Unsupervised learning (B): Useful for clustering sensor data, but again not the main paradigm.

Reinforcement learning (C): Correct, since self-driving fundamentally depends on RL decision-making.

Thus, the correct answer is Option C (Reinforcement Learning).

Images are broadly categorized based on how they store visual information:

Rasterized images (Option i):

Composed of a grid of pixels (bitmap).

Each pixel has color information.

Examples: JPEG, PNG, BMP.

Best for photos or complex visuals.

Vectorized images (Option ii):

Composed of paths defined by mathematical formulas.

Scalable without quality loss.

Examples: SVG, EPS, AI.

Best for logos, icons, and illustrations.

Sectorized images (Option iii):

Not a standard category in computer graphics.

Thus, image files are categorized into Rasterized and Vectorized, making Option A (i, ii) correct.

Inductive learning techniques in machine learning (such as decision trees, neural networks, or SVMs) follow a systematic sequence of steps for classification:

Training Step (i): A model is built using training data, where the system learns relationships between features and target labels.

Test Step (ii): The trained model is evaluated on unseen test data to measure its performance and generalizability.

Validation Step (iii): Often, a validation set is used to fine-tune model parameters, avoid overfitting, and choose the best model configuration.

Application Step (iv): The final validated model is applied to classify new, real-world data.

Since all four steps (i, ii, iii, iv) are essential to inductive classification, the correct answer is Option D (i, ii, iii, iv).