Data Science Fundamentals

Data Science…

• uses machine learning algorithms to discover patterns in data that are not immediately apparent.

• obtains meaningful insights and valuable information from data using statistical analysis, programming, and domain knowledge.

• draws from the formal sciences including hypothesis creation and testing and formation of repeatable processes.

• all of the above.

Answer explanation

Data science is a formal, applied, and interdisciplinary science that works with and analyzes large amounts of data to provide meaningful information that can be used to make decisions and solve problems. Data science includes work in computation, statistics, analytics, data mining, and programming.

The purpose of data science is:

(select all that apply)

• to analyze large amounts of structured or unstructured data to discover meaningful patterns.

• to make evidence-based recommendations to improve a process or add value to a business.

• to create a machine learning model as a goal for every project.

• to better understand a dataset through exploratory data analysis including both data visualization and statistical tests.

• to communicate the story of data to a target audience in a way that motivates action.

Answer explanation

Data Science may provide information that can be acted upon, explanations that can be tested, and predictions that are useful. It can inform recommendations to improve a process or organization if the story of the data is communicated to a target audience in an effective way. A ML model may not always be a part of your project.

An ____________ is a series of repeatable steps or rules used to accomplish specific data science tasks or solve a problem.

Answer explanation

Algorithms are the foundation of AI, ML, and DL. Simply put, recipes for computers.

Data scientists need to understand certain algorithms in so far as to be able to identify how appropriate an algorithm is for a particular problem as well as to be able to effectively tune the model.

"_____________ _______________ is the science and art of programming computers so they can learn from data." -Aurélien Géron

Answer explanation

Machine Learning helps computers predict outcomes without explicit human input; it's the process of 'learning' the underlying patterns in a set of observations that is represented through data. Patterns are extracted using algorithms and can then be represented in a mathematical model. The model is then used to predict outcomes of new observations.

What is the difference between Artificial Intelligence and Machine Learning?

(select all that apply)

• Machine learning is a subset of artificial intelligence.

• Artificial Intelligence is a branch of machine learning

• Machine learning takes in a training data set to discover patterns and create algorithms that can be applied to future data to predict outcomes.

• Artificial intelligence may not have the ability to learn or adjust based on new data but can mimic cognitive functions and perform difficult and repetitive tasks based on defined rules or algorithms.

• Machine learning uses an algorithm to identify patterns from your training dataset that it can use to define parameters for a model that can then be used to make predictions on new data.

Answer explanation

Once a machine learning program collects information and supplies outputs, AI can be used to create a new algorithm so that a machine learning program can update its model’s parameters accordingly. “AI is the effort to automate intellectual tasks normally performed by humans. Machine learning is an algorithm to discover data representation rules to execute a data-processing task, given examples of what’s expected.” - Francois Chollet

Unsupervised machine learning algorithms find connections between observations in a train dataset based only on features, without human input, and may go on to group observations based on their findings.

• True

• False

Answer explanation

Unsupervised machine learning algorithms are considered self-learning as they do not rely on labeled observations, observations with a known outcome or human input. Whereas the supervised algorithm would accept and use the labels assigned to it to model the relationship between the inputs (features) and output (target), an unsupervised algorithm would learn the differences of observations using only the features (no output) and assign its own labels to differentiate.

What are the four Vs of Data Big Data?

(select all that apply)

• Viscosity

• Volume

• Variety

• Velocity

• Veracity

Answer explanation

Big data is most commonly used to mean a massive volume of both structured and unstructured data that is so large it is difficult to process using traditional database and software techniques. The volume of big data is too big, it moves too fast, or it exceeds current processing capacity.

__________ _________ is data which does not fit a predefined data model. Often this data does not fit into the typical row-column structure of a database.

Answer explanation

Images, emails, videos, audio, and pretty much anything else that might be difficult to "tabify" might constitute examples of unstructured data.

These configuration variables are internal to a model, so the values are estimated or learned from the data when the train dataset is being fit or trained. These are the parts of the model that are learned from historical training data.

• Targets

• Hyperparameters

• Features

• Parameters

Answer explanation

The parameters customize the model to your particular problem so that you can then predict outcomes on new observations. They are saved as part of the learned model. Examples include the split points of a decision tree, the support vectors in a support vector machine, and the coefficients in a linear regression or logistic regression.

Supervised machine learning algorithms train on data sets that include the labels or outcomes for each observation.

• True

• False

Answer explanation

Supervised machine learning algorithms use labeled observations, i.e. observations with a known outcome or human input, to develop a model that will then be used to predict or estimate the outcome of new observations which are not known or labeled.