DataCamp: Data Science for Business

Question 1

What is data science

Answer 1

It’s a set of methodologies used to gather thousands of forms of data available and draw meaningful conclusions from the data.

Question 2

What can data do ?

Answer 2

describe current state of an organisation or process
detect anomalous events
diagnose the cause of events and behaviours
predict future events

Question 3

Data Science Workflow

Answer 3

Data Collection
Exploration and Visualization
Experimentation and Prediction

Question 4

Applications of Data Science:

Answer 4

Traditional Machine Learning

Internet of Things - Refers to gadgets that are not standard computers but have ability to transmit data such as smart watches and home appliances.

Image classification using Deep Learning

Question 5

Data Pseudonymization

Answer 5

Anonymization of data to protect privacy.

Question 6

PII - Personally Identifiable Information

Answer 6

Data that can be linked to an individual

Question 7

Data Sources

Answer 7

Web events, logistics data, Financial transactions, customer data

Question 8

Solicited Data (SD)

Answer 8

Obtained by requesting opinion from customers such as surveys, in-app questionnaires, focus groups and customer reviews. 
SD are used to:
1.  De-risk decision making
2. Monitor quality and
3. Create marketing collateral.

Question 9

Types of Solicited Data

Answer 9

Qualitative (subjective): conversations, open-ended questions. Help to generate hypothesis.
Quantitative: multiple choice, rating scale. Used to validate hypothesis.

Question 10

Net Promoter Score

Answer 10

Quantitative method of measuring stated preference

Question 11

Other data sources

Answer 11

APIs (a way of requesting data from 3rd parties over the internet), such as Twitter, Wikipedia, Google Maps, Yahoo Finance
Public records

Mechanical Turks - Manual data input by humans

Question 12

Unstructured data

Answer 12

such as emails, text, video and audio files, web pages and social media are stored in document databases. Use NoSQL for data retrieval

Question 13

Tabular data

Answer 13

Relational database. Use SQL for data retrieval.

Question 14

Dashboards

Answer 14

A dshboard is a set of metrics, usually in the form of graphs, that update on a pre-defined frequency such as daily, weekly, real-time etc. They help to visualize and explore collected data. Examples of dashboards include time series (which tracks a value over time)

Question 15

Ad hoc request

Answer 15

Requests for data that does not need to be repeated on a weekly or daily basis. Such request should be specific, include context and priority.

Question 16

A/B Testing

Answer 16

A/B Testing is a type of experiment for de-risking choices between two options such as changes to a website, addition of new features or wording email subjects.

Question 17

A/B Testing steps

Answer 17

1. Picking a metric to track
2. Calculating the sample size using baseline metric and test sensitivity.
3. Running the experiment
4. Determining the significance

Question 18

Machine Learning

Answer 18

A set of methods for making predictions based on existing data.

Question 19

Supervised learning

Answer 19

A subset of machine learning where the data has labels and features which are used for making predictions. Used to solve problems such as recommendation systems, email subject optimization and churn prediction.

Question 20

Clustering

Answer 20

ML algorithms that divide data into clusters. It is applied in Customer segmentation (where customers are didvided into different groups with common attributes), Image categorisation and anomaly detection. It is a part of Unsupervised learning (which uses data with only features and no labels).

Question 21

Special topics in Machine Learning

Time Series

Answer 21

Time series forecasting is any type of ML where time is an important feature. It shows periodic patterns and can help spot seasonality.

Question 22

Natural Language Processing (NLP)

Answer 22

Refers to ML problem where the dataset (input data) is text. Possible applications include customer reviews, tweet, medical records and email subjects. NLP can be used to classify sentiment and cluster medical records.

Question 23

Deep Learning (Neural Networks)

Answer 23

An area of ML used to solve more complex problems. Requires more data than traditional ML. Best used where inputs are less structured such as large amounts of texts and images. Main drawback is the lack of explainability of predictions but it makes highly accurate predictions.

Question 24

Explainable AI

Answer 24

Refers to methods that allow humans to understand the factors behind each prediction.