100+ Data Science Intern Interview Questions and Answers

The task is to rotate a square matrix by 90 degrees in an anti-clockwise direction using constant extra space.

• Iterate through each layer of the matrix from outer to inner

• For each layer, perform a four-way swap of elements

• Continue this process until all layers have been rotated

Gradient descent is an optimization algorithm used to minimize the cost function of a machine learning model.

• Gradient descent is used to update the parameters of a model to minimize the cost function.

• It follows the direction of steepest descent, which is the negative gradient of the cost function.

• The learning rate determines the step size of the algorithm.

• The formula for gradient descent is: theta = theta - alpha * (1/m) * sum((hypothesis - y) * x)

• The cost function should be ...read more

Get second highest element from an array of strings with O(N) time complexity and O(1) space complexity.

• Initialize two variables to store the highest and second highest elements.

• Traverse the array and update the variables accordingly.

• Return the second highest element.

• Handle edge cases like empty array or array with only one element.

The function finds the value in a specified categorical column with the highest average of a specified numeric column.

• Use the groupby() function in pandas to group the data by the categorical column

• Calculate the average of the numeric column for each group

• Find the group with the highest average and return the corresponding value from the categorical column

Coffiecent of x^7 in a given equation

• Use the binomial theorem to expand the equation

• Identify the term with x^7

• The coefficient of x^7 is the coefficient of that term

Random forest is an ensemble learning method that constructs a multitude of decision trees and outputs the mode of the classes. Gini impurity is a measure of impurity or randomness used in decision trees.

• Random forest is a collection of decision trees that are trained on different subsets of the data.

• Each decision tree in the forest is trained on a random subset of the features.

• The final prediction is made by taking the mode of the predictions of all the trees.

• Gini impurity i...read more

I used the Random Forest algorithm in my project.

• Random Forest is an ensemble learning method that combines multiple decision trees to make predictions.

• It is used for both classification and regression tasks.

• Random Forest reduces overfitting and provides feature importance.

• Example: I used Random Forest to predict customer churn in a telecom company.

Call by value passes a copy of the value while call by reference passes the address of the value.

• Call by value passes a copy of the value while call by reference passes the address of the value.

• Call by value does not modify the original value while call by reference can modify the original value.

• Call by value is used for simple data types while call by reference is used for complex data types.

Choosing the optimal K value in K-means clustering is crucial for accurate results.

• Elbow method: Plotting the sum of squared distances vs. K and selecting the K value where the curve bends like an elbow.

• Silhouette method: Calculating the average silhouette score for different K values and choosing the one with the highest score.

• Gap statistic method: Comparing the within-cluster dispersion to a reference null distribution to find the optimal K value.

• Cross-validation: Splitting...read more

Find students who scored more than avg marks in both 11th and 12th grades.

• Calculate the average marks for each student in 11th and 12th grades.

• Compare each student's marks with the respective average marks to find those who scored higher in both grades.

Use a SQL query to find customers who have ordered all products from all categories.

• Join the Customers, Orders, and Products tables

• Group by customer and count the distinct products ordered

• Filter for customers who have ordered the total number of products available in each category

Joins in SQL are used to combine rows from two or more tables based on a related column between them.

• Joins are used to retrieve data from multiple tables in a single query.

• Common types of joins include inner join, left join, right join, and full outer join.

• Joins are performed using the JOIN keyword and specifying the columns to join on.

• Joins can be used to combine tables based on matching values or non-matching values.

• Joins help in creating relationships between tables and fe...read more

XGBoost is preferred over Random Forest for low bias models due to its ability to reduce bias further.

• XGBoost is a more complex algorithm compared to Random Forest, allowing it to reduce bias further in low bias models.

• XGBoost uses gradient boosting which helps in reducing bias by optimizing the loss function iteratively.

• Random Forest may not be able to further reduce bias in low bias models as effectively as XGBoost.

• In scenarios where the model already has low bias, XGBoost'...read more

Overfitting is when a model is too complex and fits the training data too closely, leading to poor performance on new data.

• Regularization techniques like L1 and L2 can be used to prevent overfitting

• Cross-validation can be used to evaluate model performance on new data

• Reducing the complexity of the model can also help prevent overfitting

• Handling missing values can be done by imputing them with mean, median or mode values

• Alternatively, missing values can be dropped if they are ...read more

SQL is a programming language used for managing and manipulating relational databases. A database is a structured collection of data.

• SQL is used to retrieve, insert, update, and delete data from a database.

• A database is a software system that stores and organizes data in a structured manner.

• SQL allows users to define the structure of a database, create tables, and establish relationships between tables.

• Examples of databases include MySQL, Oracle, and SQL Server.

The average number of tea people drink in a day in Delhi varies depending on individual preferences and habits.

• The average number of tea consumed can range from 1-5 cups per day.

• Factors such as age, gender, occupation, and cultural background can influence tea consumption.

• Some people may drink more tea in the morning for a caffeine boost, while others may prefer tea throughout the day for relaxation.

• Tea consumption may also vary based on the season, with more tea being consum...read more

I chose data science because of its potential to solve complex problems and make meaningful insights from data.

• Fascination with the power of data to drive decision-making

• Interest in solving real-world problems using data-driven approaches

• Passion for exploring patterns and trends in data

• Desire to contribute to advancements in technology and innovation

• Excitement about the interdisciplinary nature of data science

• Examples: Predictive analytics in healthcare, fraud detection in fi...read more

Using linear regression for classification can lead to inaccurate predictions and unreliable results.

• Linear regression assumes a continuous output, making it unsuitable for discrete classification tasks.

• It may not handle outliers well, leading to incorrect classification boundaries.

• The predicted values may fall outside the 0-1 range for binary classification.

• Logistic regression is a more appropriate choice for classification tasks.

Mutable data structures can be modified after creation, while immutable data structures cannot be changed once created.

• Mutable data structures allow for in-place modifications, while immutable data structures require creating a new instance when modifications are needed.

• Examples of mutable data structures include lists, dictionaries, and sets in Python.

• Examples of immutable data structures include tuples and strings in Python.

Regression model is a ML algorithm used to predict continuous numerical values. Intercept is the value of the predicted variable when all predictors are zero. Coefficient represents the change in the predicted variable for a one-unit change in the predictor variable.

• Regression model predicts continuous numerical values.

• Intercept is the value of the predicted variable when all predictors are zero.

• Coefficient represents the change in the predicted variable for a one-unit change...read more

Handle missing data in time series data by imputation, interpolation, or deletion.

• Impute missing values using mean, median, mode, or predictive modeling techniques.

• Interpolate missing values using linear interpolation, spline interpolation, or time-based interpolation.

• Delete rows with missing values if they are few and do not significantly impact the analysis.

Linear regression is a statistical method to model the relationship between a dependent variable and one or more independent variables. Logistic regression is used to model the probability of a binary outcome.

• Linear regression is used for predicting continuous outcomes, while logistic regression is used for predicting binary outcomes.

• Linear regression assumes a linear relationship between the independent and dependent variables, while logistic regression uses a logistic funct...read more

Supervised learning uses labeled data to train a model, while unsupervised learning uses unlabeled data.

• Supervised learning requires a target variable for training

• Examples of supervised learning include regression and classification tasks

• Unsupervised learning finds patterns in data without predefined labels

• Examples of unsupervised learning include clustering and dimensionality reduction

A dictionary sorted in ascending order based on keys.

• Create a dictionary with key-value pairs

• Use the sorted() function to sort the dictionary based on keys

• Convert the sorted dictionary into a list of tuples

• Use the dict() constructor to create a new dictionary from the sorted list of tuples

HAVING clause is used with GROUP BY to filter grouped rows, WHERE clause is used to filter individual rows.

• HAVING clause is used with GROUP BY to filter grouped rows based on aggregate functions

• WHERE clause is used to filter individual rows based on conditions

• HAVING clause is applied after GROUP BY, WHERE clause is applied before GROUP BY

• HAVING clause can only be used with SELECT statement that contains a GROUP BY clause

Object-oriented programming (OOP) is a programming paradigm that uses objects to represent and manipulate data.

• OOP is based on the concept of classes and objects.

• Encapsulation, inheritance, and polymorphism are key principles of OOP.

• Examples of OOP languages include Java, C++, and Python.

Digital Image Processing is the manipulation of digital images using algorithms to enhance, analyze, or extract information.

• It involves techniques like image enhancement, restoration, segmentation, and compression.

• Common applications include medical imaging, satellite imaging, facial recognition, and object detection.

• Algorithms like edge detection, noise reduction, and image stitching are used in digital image processing.

• It plays a crucial role in fields like computer vision,...read more

Recommendation, personalization, fraud detection, search algorithms are used in e-commerce companies.

• Recommendation algorithms suggest products based on user behavior and preferences.

• Personalization algorithms customize the user experience based on their past behavior.

• Fraud detection algorithms identify and prevent fraudulent transactions.

• Search algorithms help users find products based on their search queries.

• Clustering algorithms group similar products together for easier n...read more

Check if a number is a palindrome or not

• Convert the number to a string

• Reverse the string and compare it with the original string

• If they are the same, the number is a palindrome

Good and bad aspects of Data Science

• Good: Data science helps in making informed decisions based on data-driven insights

• Good: Data science can uncover valuable patterns and trends in large datasets

• Bad: Data science can be time-consuming and resource-intensive

• Bad: Data science may face challenges with data privacy and ethical considerations

I tackle pressurized situations by staying calm, prioritizing tasks, seeking help when needed, and maintaining a positive attitude.

• Stay calm and composed under pressure

• Prioritize tasks based on urgency and importance

• Seek help or guidance from colleagues or supervisors

• Maintain a positive attitude and focus on finding solutions

I am familiar with a variety of technologies commonly used in data science, including programming languages, databases, and machine learning tools.

• Programming languages: Python, R, SQL

• Databases: MySQL, MongoDB

• Machine learning tools: TensorFlow, scikit-learn

Central Limit Theorem states that the sampling distribution of the sample mean approaches a normal distribution as the sample size increases.

• Central Limit Theorem is used to make inferences about a population mean based on the sample mean.

• It allows us to use the properties of the normal distribution to estimate population parameters.

• It is essential in hypothesis testing and constructing confidence intervals.

• For example, if we take multiple samples of a population and calculat...read more

Data science is a field that uses scientific methods, algorithms, and systems to extract knowledge and insights from structured and unstructured data.

• Data science involves analyzing large amounts of data to uncover patterns, trends, and insights.

• It combines statistics, machine learning, and domain knowledge to solve complex problems.

• Data science is used in various industries such as healthcare, finance, marketing, and more.

• Examples include predicting customer behavior, optimi...read more

Formulas for Precision, Recall, Accuracy, F1 Score in data science.

• Precision = TP / (TP + FP)

• Recall = TP / (TP + FN)

• Accuracy = (TP + TN) / (TP + TN + FP + FN)

• F1 Score = 2 * (Precision * Recall) / (Precision + Recall)

Overfitting and underfitting are common issues in machine learning where the model either learns the noise in the training data or fails to capture the underlying patterns.

• Overfitting occurs when a model learns the training data too well, including noise and outliers, leading to poor generalization on new data.

• Underfitting happens when a model is too simple to capture the underlying patterns in the data, resulting in high bias and low variance.

• Examples of overfitting include ...read more

A credit card company is a financial institution that issues credit cards to consumers for making purchases and borrowing money.

• Credit card companies issue credit cards to consumers for making purchases and borrowing money

• They charge interest on outstanding balances and fees for late payments

• They provide customer service for cardholders and handle disputes and fraud claims

• Examples include Visa, Mastercard, American Express, and Discover

I have used techniques like one-hot encoding, feature scaling, polynomial features, and interaction terms for feature engineering.

• One-hot encoding for categorical variables

• Feature scaling for numerical variables

• Polynomial features for capturing non-linear relationships

• Interaction terms for capturing interactions between features

Typing google.com sends a request to Google's servers, which respond with the website's HTML code.

• The browser sends a DNS request to resolve the domain name 'google.com' to an IP address

• The browser establishes a TCP connection with Google's server

• The browser sends an HTTP request to the server for the website's HTML code

• The server responds with the HTML code, which the browser renders as a webpage

Time series data can be analyzed using techniques like smoothing, decomposition, forecasting, and anomaly detection.

• Use smoothing techniques like moving averages to remove noise and identify trends.

• Apply decomposition methods like seasonal decomposition of time series (STL) to separate trend, seasonality, and residual components.

• Utilize forecasting models such as ARIMA, SARIMA, or Prophet to predict future values.

• Implement anomaly detection algorithms like Isolation Forest or...read more

Sort nearly sorted array using min heap

• Create a min heap of size k+1

• Insert first k+1 elements into min heap

• For remaining elements, extract min and insert new element

• Extract all remaining elements from min heap

• Time complexity: O(nlogk)

• Example: ['apple', 'banana', 'cherry', 'date', 'elderberry']

GAN stands for Generative Adversarial Network, a type of neural network used for generating new data.

• Consists of two neural networks - generator and discriminator

• Generator creates new data samples while discriminator tries to distinguish between real and generated data

• Used in image generation, text generation, and other creative applications