ML

Question 1

What is the main purpose of a dashboard in an ML project?

Answer 1

To display key performance indicators in a graphical user interface

Dashboards utilize data visualization and are linked to the data source for interactivity.

Question 2

True or false: A dashboard is primarily a backend application on a server.

Answer 2

FALSE

A dashboard is a graphical user interface, while an API is the backend application.

Question 3

What does an API provide for clients in an ML project?

Answer 3

An endpoint URL for data input and prediction output

The API sends responses typically in JSON format.

Question 4

Fill in the blank: An API exposes an endpoint URL which the frontend app can make a _______ request to send data.

Answer 4

HTTP POST

This allows communication between the frontend client app and the backend API.

Question 5

What are the common HTTP status codes returned by an API?

Answer 5

• 200 range: Successful operation
• 400 range: Client error
• 500 range: Server error

These codes help indicate the status of the API request.

Question 6

What is a key challenge when conveying ML results to a non-technical audience?

Answer 6

Making complex statistical results understandable

Interactive visualizations, like dashboards, can help bridge this gap.

Question 7

What is the difference between a dashboard and an API?

Answer 7

• Dashboard: Standalone solution, graphical/tabular format
• API: Integrated into existing client frontend, supplies data as JSON

Dashboards are great for explaining findings, while APIs are more production-ready.

Question 8

What is required for authentication when accessing an API?

Answer 8

An API key in the authorization header

Clear instructions must be provided for customers to access the API.

Question 9

What is a simple framework for creating an API?

Answer 9

Flask

For more complicated production-ready versions, Django DRF can be used.

Question 10

What does a 400 BAD REQUEST error indicate?

Answer 10

Incorrect data sent that the server did not understand

This is a common client error returned by an API.

Question 11

What is the role of error messages in an API?

Answer 11

To provide appropriate codes and consistent formats for client display

This helps clients understand issues that arise during API requests.

Question 12

What is the purpose of a dashboard in data analysis?

Answer 12

To explain findings and assess if the proposed hypothesis yielded expected predictions and met business requirements

A dashboard serves as a storytelling tool for a non-technical audience.

Question 13

What should be included in a dashboard summary?

Answer 13

• Problem statement
• Business requirements
• Associated jargon

This summary helps clarify the context for the audience.

Question 14

What are the two key aspects of dashboard design?

Answer 14

• Art
• Science

Effective dashboard design combines aesthetic appeal with clear communication of complex ideas.

Question 15

What should be prioritized in a dashboard to answer business questions?

Answer 15

The most important data

Prioritization ensures clarity and relevance in the information presented.

Question 16

What type of controls should be used for a small number of categorical classes?

Answer 16

• Dropdown
• Radio button

These controls allow users to select from predefined options easily.

Question 17

What control should be used for a continuous numerical feature?

Answer 17

Slider

A slider allows users to adjust values within a range intuitively.

Question 18

What is the principle of progressive disclosure in dashboard design?

Answer 18

Don’t show everything at once; let the user drill down

This approach helps manage complexity and enhances user experience.

Question 19

Why is it important to accentuate visuals in a dashboard?

Answer 19

With titles and labels

Clear titles and labels help users understand the data being presented.

Question 20

What are the different types of users that may interact with a dashboard?

Answer 20

• Analyst
• Manager
• ML Engineer
• Data Scientist

Different users have varying needs for data access and interaction.

Question 21

What might be necessary for different user types regarding dashboard access?

Answer 21

Authentication with different access levels

This ensures that users only see data relevant to their role.

Question 22

Does Streamlit support native authentication?

Answer 22

No

However, dashboards can be built with Flask or Django to implement authentication.

Question 23

What does authentication also apply to besides dashboards?

Answer 23

APIs

Ensuring secure access is crucial for both dashboards and APIs.

Question 24

What is Artificial Intelligence?

Answer 24

Simulation of human intelligence and mimicry of the human brain’s problem-solving and decision-making skills

AI mimics the brain using algorithms that adapt to input using mathematical functions.

Question 25

True or false: AI is good at solving a specific problem based on data but finds it easier to transfer techniques to similar problems.

Answer 25

FALSE ## Footnote AI struggles to transfer techniques to similar problems.

Question 26

What is the sub-field of AI that deals with how computers learn?

Answer 26

Machine learning ## Footnote Machine learning uses large datasets processed and labeled by humans.

Question 27

In machine learning, what are **labels** or **targets**?

Answer 27

Synonyms for the variable you are interested in predicting ## Footnote Examples include categorical variables like 'car' or 'no car'.

Question 28

What is a **classifier** in machine learning?

Answer 28

An estimator fitted to the model that classifies response variables into predefined classes ## Footnote It can be used for binary or multiple classifications.

Question 29

What does **regression** refer to in machine learning?

Answer 29

The iterative process of fitting a line to a numeric data set ## Footnote It predicts future numbers based on past data.

Question 30

Fill in the blank: A **clustering task** is where you want to group data but don’t have predefined _______.

Answer 30

classes ## Footnote An example is analyzing past customer purchase patterns.

Question 31

What is the purpose of **model optimizing**?

Answer 31

To reduce discrepancies between known and model estimates ## Footnote It applies weighting to improve prediction accuracy.

Question 32

What is an **algorithm** in the context of machine learning?

Answer 32

A set of instructions used to analyze input data and predict output values ## Footnote Algorithms learn without explicit programming as they are exposed to more data.

Question 33

What is a **prediction** in machine learning?

Answer 33

The output provided by a trained machine learning model based on input data ## Footnote Predictions can be used in decision-making processes or as inputs for other systems.

Question 34

What are **features** in a dataset?

Answer 34

Factors on which the response or prediction variable depends ## Footnote Features are also known as attributes.

Question 35

What is the difference between **features** and **attributes**?

Answer 35

Features refer to form, shape, or proportion; attributes refer to characteristics or qualities ## Footnote Both are used to describe data in machine learning.

Question 36

What is an **instance** in a dataset?

Answer 36

A single data point or a sample ## Footnote It can also be referred to as an example or observation.

Question 37

What is the role of a **data practitioner** in machine learning?

Answer 37

To help refine algorithms and reduce errors in model predictions ## Footnote They work with data to make predictions or classify items.

Question 38

True or false: Machine learning models can fit in a multidimensional plane for complex data.

Answer 38

TRUE ## Footnote This allows for more sophisticated predictions.

Question 39

What is the primary goal of **machine learning**?

Answer 39

To gain new abilities or knowledge through the analysis of prior data ## Footnote This involves utilizing various methods to analyze and interpret data.

Question 40

What type of learning is most commonly used in practical machine learning use cases?

Answer 40

Supervised learning ## Footnote It is often the first method encountered in workplace applications.

Question 41

In supervised learning, what are the two types of variables in the training dataset?

Answer 41

* Features * Label variable (target) ## Footnote Features are the input variables, while the label is what we aim to predict.

Question 42

In the context of supervised learning, what is the **label**?

Answer 42

The variable we aim to predict ## Footnote In the provided example, the label is the Review.

Question 43

What is the process of mapping the relationship between features and the label in supervised learning called?

Answer 43

Training the algorithm ## Footnote This involves finding patterns in the training data.

Question 44

What is the term for the task in supervised learning when the target variable is a class or category?

Answer 44

Classification ## Footnote An example is determining if an image contains a car.

Question 45

What is an example of a **regression** task in supervised learning?

Answer 45

Predicting waiter tip values in a restaurant ## Footnote Regression tasks involve predicting continuous values.

Question 46

What does **unsupervised learning** involve?

Answer 46

Only features provided without any label or target ## Footnote The algorithm discovers patterns in the data without correct answers.

Question 47

What is an example of **clustering** in unsupervised learning?

Answer 47

Grouping data by similarity ## Footnote This helps in identifying patterns without predefined labels.

Question 48

What is the purpose of **association rules** in unsupervised learning?

Answer 48

To find relationships between variables ## Footnote An example is determining if buying X leads to buying Y.

Question 49

Why is unsupervised learning important in today's technological world?

Answer 49

Computers generate vast amounts of unlabelled data ## Footnote Labeling all data in advance is often impractical.

Question 50

What is the **caveat** of unsupervised learning?

Answer 50

No guarantee that the model will output relevant results ## Footnote The algorithm works based on its internal rules without true answers.

Question 51

What is **semi-supervised learning**?

Answer 51

A mix of labeled and unlabeled data ## Footnote It occurs when some data points are labeled while others are not.

Question 52

What is a potential option when dealing with semi-supervised learning?

Answer 52

* Drop unlabeled examples and use supervised learning * Predict labels for unlabeled examples * Use only input variables in an unsupervised model ## Footnote Each option has its own implications for model performance.

Question 53

What is the main concept of **reinforcement learning**?

Answer 53

An agent learns through punishment or reward in an environment ## Footnote The agent discovers which actions lead to achieving its goal.

Question 54

In reinforcement learning, what does the **state** represent?

Answer 54

The current world or environment of the task ## Footnote It is crucial for the agent's decision-making process.

Question 55

What is an example of reinforcement learning in action?

Answer 55

An agent playing noughts and crosses (tic-tac-toe) ## Footnote The agent learns from sequences of actions and their outcomes.

Question 56

What are some specialized methods in machine learning that were mentioned but not detailed?

Answer 56

* Self-supervised * Multi-instance * Inductive * Deductive * Transductive * Multi-task * Active * Online * Transfer * Ensemble ## Footnote These methods may appear in literature or workplace applications.

Question 57

What is the overall purpose of learning methods in machine learning?

Answer 57

To analyze large quantities of data and identify patterns ## Footnote This capability exceeds human analytical abilities.

Question 58

What is a **machine learning model** in abstract form?

Answer 58

A recipe or set of instructions that takes inputs and returns an output ## Footnote It learns patterns from historical data to make predictions on new unseen data.

Question 59

What does an **ML model** generate to learn patterns?

Answer 59

Algorithms ## Footnote The type of algorithm used will impact the result.

Question 60

In a scatter plot for a dataset of Paw and Weight measurements, what does the **x-axis** represent?

Answer 60

Paw Size ## Footnote The y-axis represents weight.

Question 61

What does the **line** in a scatter plot represent in the context of a machine learning model?

Answer 61

The output of the algorithm ## Footnote It indicates how the ML model makes predictions.

Question 62

What are the two parts to consider about the graph in a machine learning model?

Answer 62

* The line itself (straight or squiggly) * Its placement in the graph ## Footnote These aspects visually represent how the model classifies data.

Question 63

What is an **ML pipeline**?

Answer 63

A sequence of tasks performed when training a machine learning model ## Footnote Typical tasks include data validation, model training, model tuning, and model deployment.

Question 64

What is the last step in an ML pipeline?

Answer 64

The ML model ## Footnote The preceding steps prepare the data for the model.

Question 65

What are the **three commonly used machine learning tasks**?

Answer 65

* Regression * Classification * Clustering ## Footnote Regression and Classification are used in supervised learning, while Clustering is used in unsupervised learning.

Question 66

When would you use a **regression problem**?

Answer 66

When the target variable is a continuous numerical value ## Footnote This often involves complex relationships between multiple attributes.

Question 67

What type of problem is it if the label variable you wish to predict is a **class**?

Answer 67

Classification problem ## Footnote Classes can be strings, numbers, or Booleans, and must be nominal or ordinal values.

Question 68

What is **multi-class classification**?

Answer 68

Classification problem with more than 2 classes ## Footnote The type of classification algorithm used depends on the number of classes.

Question 69

What is the main type of **unsupervised learning algorithm**?

Answer 69

Clustering ## Footnote It groups data points based on their similarities.

Question 70

What does clustering do with data points?

Answer 70

Groups them based on similar properties ## Footnote Dissimilar data points are placed in separate clusters.

Question 71

What is **Scikit-learn**?

Answer 71

A free open-source machine learning library written in Python ## Footnote It contains tools for predictive data analysis and is built upon NumPy, SciPy, and Matplotlib.

Question 72

What is the common alias used to import **Scikit-learn** in Python scripts?

Answer 72

sklearn ## Footnote This alias is commonly used for convenience in coding.

Question 73

List the main functionalities provided by **Scikit-learn**.

Answer 73

* Classification * Regression * Clustering * Data processing * Dimensionality reduction * Feature engineering * Feature scaling * Feature selection * Tuning model hyperparameters * Creating ML pipelines * Evaluating model performance ## Footnote These functionalities make it a comprehensive tool for machine learning tasks.

Question 74

Define an **estimator** in the context of Scikit-learn.

Answer 74

A mathematical function that learns from data ## Footnote Examples include machine learning models like random forests or linear regression.

Question 75

What are the two types of **estimators** in Scikit-learn?

Answer 75

* Predictors * Transformers ## Footnote Each type has different methods for learning and transforming data.

Question 76

What methods do **predictor estimators** use?

Answer 76

* .fit() * .predict() ## Footnote These methods allow the model to learn patterns from data and make predictions.

Question 77

What methods do **transformer estimators** use?

Answer 77

* .fit() * .transform() ## Footnote These methods allow the estimator to learn from data and transform it into a better distribution.

Question 78

What is a **pipeline** in Scikit-learn?

Answer 78

A sequence of tasks arranged into a single function ## Footnote It typically includes data preparation steps followed by the ML model.

Question 79

Why is **Scikit-learn** important for data practitioners?

Answer 79

It is a centralized and complete library for conventional ML ## Footnote It provides effective modules that assist from development through to deploying ML pipelines.

Question 80

What is the sub-field of AI that deals with how computers are made smart?

Answer 80

Machine Learning ## Footnote This field focuses on algorithms and statistical models that enable computers to perform tasks without explicit instructions.

Question 81

Which machine learning technique is essential for **binary classification problems**?

Answer 81

Logistic Regression ## Footnote Logistic regression is widely used for predicting binary outcomes.

Question 82

What is an **estimator** in Scikit-learn?

Answer 82

A mathematical function that learns from data ## Footnote Estimators are fundamental components in Scikit-learn for building machine learning models.

Question 83

What is the primary goal of **reinforcement learning**?

Answer 83

To learn from rewards and punishments ## Footnote This approach allows agents to learn optimal behaviors through trial and error.

Question 84

In the Boston housing dataset example, what is the **target variable**?

Answer 84

Median value of owner-occupied homes ## Footnote The target variable is what the model aims to predict.

Question 85

What is a **clustering task** in machine learning?

Answer 85

Grouping data without predefined classes ## Footnote Clustering is an unsupervised learning technique.

Question 86

In a classification report, what does '**recall**' indicate?

Answer 86

The sensitivity to positive instances ## Footnote Recall measures the ability of a model to find all relevant cases.

Question 87

What type of algorithm would you use if your **target variable** is a continuous numerical value?

Answer 87

Regression ## Footnote Regression algorithms predict continuous outcomes.

Question 88

What are the **features** within a dataset?

Answer 88

The factors on which the response variable depends ## Footnote Features are the input variables used for making predictions.

Question 89

Which statistical method is commonly used in machine learning for making **predictions**?

Answer 89

Regression ## Footnote Regression analysis is a key technique for prediction.

Question 90

What is the typical first step in a **supervised learning workflow**?

Answer 90

Split the dataset into a train and test set ## Footnote This step is crucial for evaluating model performance.

Question 91

What is the role of a **classifier** in machine learning?

Answer 91

To classify response variables into predefined classes ## Footnote Classifiers are used in supervised learning to categorize data.

Question 92

Which of these abilities does AI attempt to mimic from the **human brain**?

Answer 92

Reasoning ## Footnote AI systems aim to replicate cognitive functions.

Question 93

Which type of learning involves training a model with **labelled data**?

Answer 93

Supervised Learning ## Footnote This approach uses input-output pairs for training.

Question 94

Which metric is used to measure the performance of a **regression model**?

Answer 94

R-squared Score ## Footnote R-squared indicates the proportion of variance explained by the model.

Question 95

Which term refers to **grouping data points** based on their similarities without predefined labels?

Answer 95

Clustering ## Footnote Clustering is a fundamental concept in unsupervised learning.

Question 96

What is the primary objective of an **AI system**?

Answer 96

Mimic human intelligence ## Footnote AI aims to replicate cognitive functions and decision-making processes.

Question 97

What is the primary focus of **linear regression**?

Answer 97

Modelling the relationship between a dependent variable and one or more independent variables ## Footnote Linear regression is used to predict outcomes based on linear relationships.

Question 98

In the context of machine learning, what does a **pipeline** allow you to do?

Answer 98

Train a model, evaluate its performance, and use the model for prediction ## Footnote Pipelines streamline the workflow in machine learning projects.

Question 99

What is the term for the variable that you are interested in **predicting** in machine learning?

Answer 99

Label ## Footnote Labels are the outcomes that models aim to predict.

Question 100

In the context of AI, what is a **Turing test** used to determine?

Answer 100

Whether a system acts like a human ## Footnote The Turing test assesses a machine's ability to exhibit intelligent behavior.

Question 101

What are the **two main learning styles** in machine learning?

Answer 101

* Supervised learning * Unsupervised learning ## Footnote Supervised learning involves labeled data, while unsupervised learning works with unlabeled data.

Question 102

What is the **formula** for Linear Regression?

Answer 102

Y = mX + c ## Footnote In ML terms: β₀ (beta zero) is the intercept (c) and β₁ (beta one) is the slope (m).

Question 103

What is **Linear Regression** used for?

Answer 103

* Predicting a number * Examples: House prices, Revenue, Temperature, Sales ## Footnote It draws a straight line that best fits the data and minimizes error.

Question 104

What is the key difference between **Linear Regression** and **Logistic Regression**?

Answer 104

Logistic Regression outputs a probability (0 to 1), not a raw number ## Footnote Logistic function produces an S-shaped curve and is used for binary classification.

Question 105

What are the **examples** of Logistic Regression?

Answer 105

* Pass / Fail * Spam / Not spam * Yes / No ## Footnote Logistic regression predicts probability, then you choose how to classify it.

Question 106

What do **Decision Trees** resemble?

Answer 106

A flowchart of yes/no questions ## Footnote Each question splits the data, starting with general questions and moving toward specific decisions.

Question 107

What is a common risk associated with **Decision Trees**?

Answer 107

Overfitting ## Footnote This occurs when there are too many questions, causing the model to memorize training data.

Question 108

What is the purpose of **Ensemble Methods**?

Answer 108

* More accurate * More robust * Reduce overfitting ## Footnote Ensemble methods combine multiple models to improve predictions.

Question 109

How does **Bagging** (Bootstrap Aggregation) work?

Answer 109

* Random samples of data → many trees * Trees run in parallel * Final result = majority vote / average ## Footnote Random Forest is the most famous bagging method.

Question 110

What is a key strength of **Bagging**?

Answer 110

Reduces overfitting ## Footnote It is also fast due to parallel processing.

Question 111

How does **Boosting** differ from Bagging?

Answer 111

Trees run one after another ## Footnote Each new tree focuses on correcting previous errors, using the same dataset modified each time.

Question 112

What are the **pros** and **cons** of Boosting?

Answer 112

* Pros: Very strong predictive power * Cons: Slower (serial), more prone to overfitting if overused ## Footnote Examples include AdaBoost, Gradient Boosting, XGBoost.

Question 113

What is the main use of **K-Means** in machine learning?

Answer 113

Grouping similar data ## Footnote It is an unsupervised learning method that requires no labels.

Question 114

What are the **key assumptions** of K-Means?

Answer 114

* Points close together are similar * Clusters are roughly circular ## Footnote K-Means is sensitive to outliers and initial cluster positions.

Question 115

What is the **output type** of Linear Regression?

Answer 115

Number ## Footnote It is used to predict values.

Question 116

What is the **output type** of Logistic Regression?

Answer 116

Class (2) ## Footnote It is used for binary classification.

Question 117

What is the **output type** of Decision Trees?

Answer 117

Class / Number ## Footnote They are used for rule-based decisions.

Question 118

What is the **output type** of Random Forest (Bagging)?

Answer 118

Class / Number ## Footnote It provides strong general predictions.

Question 119

What is the **output type** of Boosting?

Answer 119

Class / Number ## Footnote It is used for high accuracy models.

Question 120

What is the **output type** of K-Means?

Answer 120

Clusters ## Footnote It is used for grouping data.

Question 121

What is the key question when choosing an **algorithm**?

Answer 121

What are you trying to predict (if anything)? ## Footnote This question guides the selection of the appropriate algorithm based on the prediction goal.

Question 122

When should you use **regression** algorithms?

Answer 122

* Predicting a continuous number * Examples: Sales revenue, House prices, Demand forecasting, Temperature ## Footnote Use regression when the output is a number, not a category.

Question 123

What are suitable algorithms for **regression**?

Answer 123

* Linear Regression * Decision Trees * Random Forest * Ensemble Tree Algorithms ## Footnote These algorithms are effective for predicting continuous numerical outcomes.

Question 124

When should you use **classification** algorithms?

Answer 124

* Predicting a category * Examples: Fraud / Not fraud, Pass / Fail, Customer churn / No churn ## Footnote Classification is used when the output is a category rather than a number.

Question 125

What are the types of **classification** and their suitable algorithms?

Answer 125

* Binary classification: Logistic Regression, Decision Trees, Random Forests, Ensemble Trees * Multi-class classification: Decision Trees, Random Forests, Ensemble Trees, Artificial Neural Networks * Image classification: Convolutional Neural Networks (CNNs) ## Footnote The choice of algorithm depends on the number of classes and the data type.

Question 126

What is the purpose of **clustering** in unsupervised learning?

Answer 126

* No labels * Want to group similar data ## Footnote Clustering is used when there is no target variable and the goal is to discover structure in the data.

Question 127

What is a common algorithm used for **clustering**?

Answer 127

K-Means ## Footnote K-Means is widely used for grouping similar data points without predefined outcomes.

Question 128

What are the **algorithm families** based on ML tasks?

Answer 128

* Predict a number: Regression (Linear regression, Trees) * Predict a class: Classification (Logistic regression, Trees, ANN) * Group data: Clustering (K-Means) * Images: Classification (CNNs) * Text / language: NLP ## Footnote This framework helps in understanding the types of algorithms based on the task.

Question 129

Why does **data quantity** matter in machine learning?

Answer 129

* Algorithms need enough data to learn patterns * Rough guidelines: Fewer than ~100 images → very hard to learn patterns; < 50–100 rows for tabular data → most models struggle ## Footnote More complex data requires more data for effective learning.

Question 130

What are **quick wins** in machine learning?

Answer 130

* Help Desk Automation (Chatbots) * Recommender Systems * Semantic Search * Using Existing Analytics as ML Opportunities * ML is often R&D ## Footnote Quick wins are low effort, high impact ML projects that can improve efficiency.

Question 131

What is the benefit of **chatbots** in help desk automation?

Answer 131

* Handles FAQs * Faster response * Frees human agents for complex issues ## Footnote Users prefer fast answers over human interaction for simple tasks.

Question 132

What is the purpose of **semantic search**?

Answer 132

* Understands meaning, not just words * Learns relationships from past searches ## Footnote This prevents duplicate documents and wasted time by providing more relevant search results.

Question 133

What can traditional reports show that may lead to **ML opportunities**?

Answer 133

* Trends * Bottlenecks * Clear targets to predict ## Footnote Defining a target variable can help uncover patterns that analytics may have missed.

Question 134

What is a key takeaway regarding choosing algorithms in machine learning?

Answer 134

* Depends on what you’re predicting * Type of output * Data type * Amount of data ## Footnote Quick ML wins usually come from automation, recommendations, search improvements, and addressing existing business pain points.