Mockpreps
🧑💼
🪴
Login now and master your exams with Mockpreps’ mock tests!
Top Interview Questions
Machine Learning (ML) is a branch of artificial intelligence (AI) that enables computers to learn from data and improve their performance over time without being explicitly programmed for every task. Instead of following fixed instructions, machine learning systems identify patterns in data and use those patterns to make predictions or decisions.
Machine learning is widely used in modern technology—from recommendation systems on streaming platforms like Netflix to voice assistants and fraud detection systems in banks.
At a high level, machine learning involves three main steps:
Machine learning models require large amounts of data to learn from. This data can come from:
Images
Text
Audio
Sensor readings
User interactions
The quality and quantity of data significantly affect model performance.
During training, the algorithm analyzes the data and identifies patterns. It adjusts its internal parameters to minimize errors.
For example, a model trained to recognize cats in images will learn features such as:
Shapes
Colors
Edges
Patterns
Once trained, the model can make predictions on new, unseen data. For example:
Predicting whether an email is spam
Recommending products
Recognizing speech
Machine learning is generally divided into three main types:
In supervised learning, the model is trained using labeled data. Each input has a corresponding correct output.
Examples:
Predicting house prices
Email spam detection
Common algorithms include:
Linear Regression
Decision Trees
Support Vector Machines
In unsupervised learning, the model works with unlabeled data and tries to find hidden patterns or groupings.
Examples:
Customer segmentation
Market basket analysis
Common techniques include:
Clustering (e.g., K-Means)
Dimensionality reduction (e.g., PCA)
Reinforcement learning involves an agent learning by interacting with an environment and receiving rewards or penalties.
Examples:
Game playing AI
Robotics
Autonomous vehicles
The model learns through trial and error to maximize rewards.
Features: Input variables used for prediction
Labels: Output or target variable
Example: In predicting house prices:
Features: size, location, number of rooms
Label: price
Training data is used to build the model
Testing data evaluates how well the model performs on unseen data
Overfitting: Model learns too much from training data, including noise, and performs poorly on new data
Underfitting: Model is too simple to capture patterns in the data
Common evaluation metrics include:
Accuracy
Precision and recall
F1 score
Mean squared error
Used for predicting continuous values, such as house prices.
Used for classification problems, such as spam detection.
Models decisions in a tree-like structure based on conditions.
An ensemble of multiple decision trees for improved accuracy.
Classifies data based on the closest data points.
Finds the optimal boundary between different classes.
Machine learning is used across many industries:
Disease diagnosis
Medical image analysis
Drug discovery
Fraud detection
Credit scoring
Algorithmic trading
Product recommendations
Customer behavior analysis
Companies like Amazon use machine learning to recommend products based on user preferences.
Platforms like Facebook use machine learning for:
Content recommendations
Image recognition
Targeted advertising
Autonomous vehicles
Traffic prediction
Route optimization
Streaming services like Netflix use machine learning to recommend movies and shows.
Machine learning automates repetitive tasks and decision-making processes.
Models improve as more data becomes available.
ML can analyze massive datasets efficiently.
Well-trained models can achieve high accuracy in predictions.
Poor-quality data leads to poor model performance.
ML models require large datasets for training.
Designing and tuning models can be complex.
Some models (like deep neural networks) are difficult to interpret.
Artificial Intelligence (AI) is a broad field focused on creating intelligent machines.
Machine Learning is a subset of AI that focuses on learning from data.
In simple terms:
AI is the goal, and machine learning is one of the ways to achieve it.
Popular tools and libraries include:
TensorFlow
PyTorch
Scikit-learn
Keras
These tools help developers build, train, and deploy machine learning models efficiently.
Machine learning is rapidly evolving and shaping the future of technology. Emerging trends include:
Deep learning advancements
Explainable AI (XAI)
Edge AI (running models on devices)
Integration with IoT and cloud computing
Machine learning is expected to play a major role in industries such as healthcare, finance, robotics, and autonomous systems.
Machine learning is a transformative technology that allows computers to learn from data and make intelligent decisions without explicit programming. From recommendation systems used by companies like Netflix to fraud detection in financial institutions, machine learning is deeply embedded in modern life.
With its wide range of applications, growing demand, and continuous advancements, machine learning is one of the most important technologies of the digital age. Understanding its concepts, types, and applications provides a strong foundation for anyone interested in artificial intelligence and data-driven technologies.
Answer:
Machine Learning (ML) is a field of AI that enables systems to learn patterns from data and make predictions or decisions without being explicitly programmed.
👉 Example:
Spam email detection
Recommendation systems
Answer:
Supervised Learning
Data is labeled
Example: House price prediction
Unsupervised Learning
Data is unlabeled
Example: Customer clustering
Reinforcement Learning
Learning via rewards and penalties
Example: Game AI, robotics
Answer:
| Concept | Description |
|---|---|
| AI | Broad field of intelligent machines |
| ML | Subset of AI that learns from data |
| DL | Subset of ML using neural networks |
Answer:
Training data → used to train the model
Testing data → used to evaluate performance
👉 Typically split like:
80% training
20% testing
Answer:
When a model learns the training data too well, including noise, and performs poorly on new data.
👉 Signs:
High training accuracy
Low test accuracy
Answer:
When a model is too simple to capture patterns.
👉 Results in:
Low training accuracy
Low testing accuracy
Answer:
Cross-validation
Regularization (L1, L2)
More training data
Dropout (in deep learning)
Pruning (decision trees)
Answer:
Steps to clean and prepare data before training:
Handling missing values
Encoding categorical data
Feature scaling
Removing duplicates
Answer:
| Normalization | Standardization |
|---|---|
| Scales data between 0 and 1 | Mean = 0, Std = 1 |
| Uses min-max scaling | Uses z-score |
Answer:
Creating new features or modifying existing ones to improve model performance.
👉 Example:
Extracting year from date
Combining features
Answer:
Used to predict continuous values.
Equation:
y = mx + c
👉 Example:
Predicting house prices
Answer:
Used for classification problems.
Outputs probability between 0 and 1
Uses sigmoid function
👉 Example:
Spam detection
Answer:
A tree-like model used for classification and regression.
Splits data based on conditions
Easy to interpret
Answer:
An ensemble of multiple decision trees.
👉 Benefits:
Reduces overfitting
Improves accuracy
Answer:
Classifies data based on the nearest neighbors.
Steps:
Choose K
Calculate distance
Assign majority class
Answer:
Finds the hyperplane that best separates classes.
👉 Works well for:
High-dimensional data
Classification problems
Answer:
Grouping similar data points without labels.
Answer:
Partitions data into K clusters.
Steps:
Choose K
Assign points to nearest centroid
Update centroids
Repeat
Answer:
Reducing number of features while retaining important information.
👉 Techniques:
PCA (Principal Component Analysis)
Answer:
Percentage of correct predictions.
Accuracy = (Correct Predictions / Total Predictions)
Answer:
Precision → how many predicted positives are correct
Recall → how many actual positives are captured
Answer:
Harmonic mean of precision and recall.
👉 Useful when classes are imbalanced
Answer:
A table showing:
True Positive (TP)
True Negative (TN)
False Positive (FP)
False Negative (FN)
Answer:
ROC curve plots TPR vs FPR
AUC measures model performance
Higher AUC = better model
Answer:
| Bias | Variance |
|---|---|
| Error from wrong assumptions | Sensitivity to data |
| High bias → underfitting | High variance → overfitting |
Answer:
Balancing underfitting and overfitting to achieve optimal performance.
Answer:
Technique to evaluate model by splitting data into multiple folds.
👉 Example: K-fold cross-validation
Answer:
Measures how far predictions are from actual values.
Answer:
Optimization algorithm used to minimize loss.
Steps:
Compute gradient
Update weights
Repeat
Answer:
Controls step size in gradient descent.
Too high → overshooting
Too low → slow convergence
Answer:
A model inspired by the human brain consisting of:
Input layer
Hidden layers
Output layer
Answer:
Introduces non-linearity.
Examples:
ReLU
Sigmoid
Tanh
Answer:
Algorithm used to update weights by propagating error backward.
Answer:
Remove rows
Mean/median imputation
Use algorithms that handle missing values
Answer:
Depends on:
Type of problem (classification/regression)
Dataset size
Interpretability
Accuracy requirements
Answer:
Algorithms like KNN, SVM get biased → hence scaling is required.
Answer:
| Parametric | Non-parametric |
|---|---|
| Fixed number of parameters | Flexible |
| Faster | More complex |
Answer:
Combining multiple models to improve performance.
Types:
Bagging (Random Forest)
Boosting (XGBoost)
Answer:
Sequential learning where each model improves the previous one.
Answer:
Training multiple models independently and combining results.
👉 Explain:
Problem statement
Dataset used
Algorithm used
Accuracy achieved
Challenges
👉 Focus on:
Interest in data-driven solutions
Real-world impact
Problem-solving mindset
Analytical thinking
Coding skills
Curiosity to learn
✔ Focus on:
Python basics (NumPy, Pandas)
Algorithms intuition (not just theory)
Model evaluation metrics
✔ Practice:
Simple ML projects:
House price prediction
Spam classifier
Customer segmentation
✔ Be ready to:
Explain projects clearly
Write basic pseudocode
Interpret results
Answer:
Machine Learning (ML) is a field of AI that enables systems to learn patterns from data and make predictions or decisions without being explicitly programmed.
👉 Example:
Spam email detection
Recommendation systems
Answer:
Supervised Learning
Data is labeled
Example: House price prediction
Unsupervised Learning
Data is unlabeled
Example: Customer clustering
Reinforcement Learning
Learning via rewards and penalties
Example: Game AI, robotics
Answer:
| Concept | Description |
|---|---|
| AI | Broad field of intelligent machines |
| ML | Subset of AI that learns from data |
| DL | Subset of ML using neural networks |
Answer:
Training data → used to train the model
Testing data → used to evaluate performance
👉 Typically split like:
80% training
20% testing
Answer:
When a model learns the training data too well, including noise, and performs poorly on new data.
👉 Signs:
High training accuracy
Low test accuracy
Answer:
When a model is too simple to capture patterns.
👉 Results in:
Low training accuracy
Low testing accuracy
Answer:
Cross-validation
Regularization (L1, L2)
More training data
Dropout (in deep learning)
Pruning (decision trees)
Answer:
Steps to clean and prepare data before training:
Handling missing values
Encoding categorical data
Feature scaling
Removing duplicates
Answer:
| Normalization | Standardization |
|---|---|
| Scales data between 0 and 1 | Mean = 0, Std = 1 |
| Uses min-max scaling | Uses z-score |
Answer:
Creating new features or modifying existing ones to improve model performance.
👉 Example:
Extracting year from date
Combining features
Answer:
Used to predict continuous values.
Equation:
y = mx + c
👉 Example:
Predicting house prices
Answer:
Used for classification problems.
Outputs probability between 0 and 1
Uses sigmoid function
👉 Example:
Spam detection
Answer:
A tree-like model used for classification and regression.
Splits data based on conditions
Easy to interpret
Answer:
An ensemble of multiple decision trees.
👉 Benefits:
Reduces overfitting
Improves accuracy
Answer:
Classifies data based on the nearest neighbors.
Steps:
Choose K
Calculate distance
Assign majority class
Answer:
Finds the hyperplane that best separates classes.
👉 Works well for:
High-dimensional data
Classification problems
Answer:
Grouping similar data points without labels.
Answer:
Partitions data into K clusters.
Steps:
Choose K
Assign points to nearest centroid
Update centroids
Repeat
Answer:
Reducing number of features while retaining important information.
👉 Techniques:
PCA (Principal Component Analysis)
Answer:
Percentage of correct predictions.
Accuracy = (Correct Predictions / Total Predictions)
Answer:
Precision → how many predicted positives are correct
Recall → how many actual positives are captured
Answer:
Harmonic mean of precision and recall.
👉 Useful when classes are imbalanced
Answer:
A table showing:
True Positive (TP)
True Negative (TN)
False Positive (FP)
False Negative (FN)
Answer:
ROC curve plots TPR vs FPR
AUC measures model performance
Higher AUC = better model
Answer:
| Bias | Variance |
|---|---|
| Error from wrong assumptions | Sensitivity to data |
| High bias → underfitting | High variance → overfitting |
Answer:
Balancing underfitting and overfitting to achieve optimal performance.
Answer:
Technique to evaluate model by splitting data into multiple folds.
👉 Example: K-fold cross-validation
Answer:
Measures how far predictions are from actual values.
Answer:
Optimization algorithm used to minimize loss.
Steps:
Compute gradient
Update weights
Repeat
Answer:
Controls step size in gradient descent.
Too high → overshooting
Too low → slow convergence
Answer:
A model inspired by the human brain consisting of:
Input layer
Hidden layers
Output layer
Answer:
Introduces non-linearity.
Examples:
ReLU
Sigmoid
Tanh
Answer:
Algorithm used to update weights by propagating error backward.
Answer:
Remove rows
Mean/median imputation
Use algorithms that handle missing values
Answer:
Depends on:
Type of problem (classification/regression)
Dataset size
Interpretability
Accuracy requirements
Answer:
Algorithms like KNN, SVM get biased → hence scaling is required.
Answer:
| Parametric | Non-parametric |
|---|---|
| Fixed number of parameters | Flexible |
| Faster | More complex |
Answer:
Combining multiple models to improve performance.
Types:
Bagging (Random Forest)
Boosting (XGBoost)
Answer:
Sequential learning where each model improves the previous one.
Answer:
Training multiple models independently and combining results.
👉 Explain:
Problem statement
Dataset used
Algorithm used
Accuracy achieved
Challenges
👉 Focus on:
Interest in data-driven solutions
Real-world impact
Problem-solving mindset
Analytical thinking
Coding skills
Curiosity to learn
✔ Focus on:
Python basics (NumPy, Pandas)
Algorithms intuition (not just theory)
Model evaluation metrics
✔ Practice:
Simple ML projects:
House price prediction
Spam classifier
Customer segmentation
✔ Be ready to:
Explain projects clearly
Write basic pseudocode
Interpret results
