Flashcards
Mid-Semester Exam
Cards in this deck (70)
- 1Marketing Research (Module 1.1)Required when a management decision problem is related to an organisation’s market performance.
It’s the organisational function of understanding the customer. - 2Customer Centric Marketing (Module 1.1)Defines the role of a company as creating value by satisfying the customers’ needs
- 3Click-through rate (Module 1.1)Share of recipients of an electronic message who click on a link in the message (individualised links in each message help track where a visitor came from).
- 4Marketing fatigue (Module 1.1)Consumers become less responsive to advertising messages because of the high volume of competing messages they receive.
- 5Data (Module 1.2)Data are not handed to us by nature, they are collected from various sources and shaped by human beliefs
- 6Big data (Module 1.2)Big data are characterised by the four V's: volume (large data sizes), variety (multiple formats), velocity (high-speed generation), and veracity (trustworthiness of data)
- 7Small data (Module 1.2)Data that is small enough for easy human comprehension; could be large (e.g., petabytes of data), but it does not meet the four Vs criteria for Big Data. Small data can be reasonably handled on computers used in our day-to-day lives. It usually has a very specific purpose (e.g. the data that you have on an Excel spreadsheet). More importantly, small data is capable of impacting decisions in the present.
- 8Legacy data source (Module 1.2)Traditionally, these are data that comes from a variety of sources, such as websites, click streams that are used to track user activities, emails, databases, images, and media.
- 9Correlation & Causation (Module 1.3)although data may have show a correlation between two variables, it does not imply causation.
- 10Understand the big picture and context (Module 1.3)Even though data might indicate one thing it may not always necessarily be an accurate reflection of reality. It is important for you as a data scientist to understand the big picture and the context of the problem.
- 11With big data comes big responsibilities (Module 1.3)When a company makes the decision to utilize big data analytics each process needs to be mapped out to have an intimate understanding of how the data will be used.
Know your data and know the potential uses of the data better. Don't be afraid to think outside the box but know the potential consequences. - 12Quality of insights matters more than the quantity (Module 1.3)when you use data, the quality of insights matters more than the number of insights.
- 13Analytics (Module 1.4)Analytics focuses on algorithms and analytical tools, and provides information to the decision-makers.
- 14Domain or Business knowledge (Module 1.4)Domain & business knowledge identify business problems and opportunities to use data science to gain insights from data. Domain experts are the ones who are responsible for decisions and actions. They understand the consequences of actions and should be the ones who assess the wider societal and business impacts of data science projects.
- 15Information technology (IT) (Module 1.4)IT experts focus on all aspects of data such as data administration, storage of data, technologies that are used to extract, transform and load data; and ways to improve the speed of access to data.
- 16Formulation of data science team (Module 2.1)
- 17Data science team structure (Module 2.1)1. Business/knowledge domain
2. Analytical Domain
4. Information technology (IT) - 18Champion (Module 2.1)(Business or Knowledge domain) Organizational data science is a team effort, and it typically consists of an organizational champion - the champion is a person who has a clear understanding of the business domain; knows what the problems are and identifies opportunities that can be addressed through data science.
- 19Project manager (Module 2.1)(Analytical domain)
In this domain, an organizational data science team will have a project manager. This is usually a person who has quite a bit of expertise and experience in data
science. The project manager should also have sufficient knowledge of the business domain. - 20Data scientist (Module 2.1)(Analytical domain)
Underneath the project manager, are data scientists who are good at statistics
and machine learning. This person should be comfortable with algorithm development,
modeling, and testing - 21Data engineer (Module 2.1)(Analytical domain)
The purpose of the data engineer is to manage all aspects of data such as data quality; extracting transforming and loading data (the acronym
for which is ELT). The data engineer manages the data pipeline; bringing together data from the different sources and integrating them in a format that can be used by the data scientists. - 22Data analyst (Module 2.1)(Analytical domain)
There is usually a data analyst reporting to the project manager. The data analyst
utilizes various BI tools and visualization tools to draw insights from the data. - 23IT manager (Module 2.1)(IT domain)
The IT manager is a person who
has expertise in the operational side of technologies required to support the data management and data analytics. - 24Software engineer (Module 2.1)(IT domain)
Under the IT manager, there is usually a software engineer who oversees operationalizing the machine learning models developed by the data scientists. They are also responsible for productizing and deploying data science solutions - 25Hardware engineer (Module 2.1)(IT domain)
There is also a hardware engineer who is comfortable with cloud platforms
such as Amazon Web Services (AWS), Google Cloud Platform (GCP) or Microsoft Azure platforms. The hardware engineer knows how to configure the technology infrastructure needed to store data and make data available to the data science team. - 26Decision contexts (Module 2.2)Decision contexts are characterised by a unique relationship between cause and effect requiring different responses by decision-makers that are contextually appropriate.
- 27Decision frameworks (Module 2.2)Decision frameworks can help decision makers identify which decision context they find themselves in so that they can make appropriate decisions and avoid the issues arising from misaligned management styles.
- 28Simple decision environment (Module 2.2)- Repeating patterns and consistent
- Clear cause and effect relationship
- Predictable and repeatable
- Known knowns
Sense -> Categorize -> Respond - 29Complicated decision environment (Module 2.2)- Cause and effect relationships discoverable but not immediately apparent
- Known unknowns
Sense -> Analyze -> Respond - 30Complex decision environments (Module 2.2)- No cause and effect relationships determined
- No right answers
- Many competing ideas
- Unknown unknowns
Probe -> Sense -> Respond - 31Chaotic decision environment (Module 2.2)- Unpredictable and emergent outcomes
- No clear cause and effect obvious
- Unknownables
Act -> Sense -> Respond - 32Descriptive Analytics (Module 2.3)What happened?
Examines historical data to answer questions - 33Diagnostic Analytics (Module 2.3)Why did it happen?
Identifies patters and discovers relationships in data - 34Predictive Analytics (Module 2.3)What will happen?
Uses current and historical data to predict future activity. - 35Prescriptive Analytics (Module 2.3)What do we do?
Applies rules and modeling for better decision-making. - 36Social physics (Module 2.3)A kind of new learning where the premise is that every human activity generates data that embeds information about human behaviour, which can then be analyzed to draw more informed predictions
- 37Visual Analytics (Module 2.3)Visual analytics provides the ability to combine data visualization with other analytical approaches to work with large volumes of data and make it easier for decision-makers to draw insights. Visual analytics includes capabilities for predictions, forecasting, simulation and so on; and it presents the results to decision-makers in a visually appealing manner.
- 38Deep learning (Module 2.3)This is a specialized type of artificial intelligence. There is a limitation in interpretability with deep learning and the transferability of findings is difficult. The models are usually called as black box models because their internal operations are not transparent and have a lot of open-ended inference. Deep learning requires situational and behavioural stability which means that the models cannot be applied in a different context than where it was originally trained and tested.
- 39Data Types - Notation (Module 2.4)
- 40Data Types - Terminology (Module 2.4)
- 41Categorical variables (Module 2.4)A categorical variable takes values in a set of categories where each category represents a qualitative attribute such as a person’s gender, city of birth, or nationality.
- Nominal
- Ordinal - 42Numerical variables (Module 2.4)A numerical variable takes numerical values where each number represents a quantitative measurement such as a person’s height, heart rate, or number of friends on Facebook.
- Discrete
-Continuous - 43Nominal variable (Module 2.4)A nominal variable is a categorical variable where the categories are unordered. Marital status is an example of a nominal variable, with possible categories {single, married}.
Python: Boolean or Integer of String - 44Ordinal variable (Module 2.4)An ordinal variable is a categorical variable where the order of the categories is important. For example, a student’s grade is naturally an ordinal variable with categories (ordered from lowest to highest) {FA, PS, CR, DI, HD}.
Python: Boolean or Integer - 45Discrete variable (Module 2.4)A discrete variable is a numerical variable where the possible values can be counted or listed. The list of possible values can be either finite or infinite.
Python: Integer - 46Continuous variable (Module 2..4)A continuous variable is a numerical variable where the possible values cannot be listed and can only be described using intervals.
Python: Float - 47Common Data Formats (Module 2.4)Commonly datasets will be stored in a database system. Let's assume that the dataset has been extracted from the database into a portable or "serialised" format.
- Microsoft Excel
- Comma separated value (CSV)
- JavaScript Object Notation (JSON) - 48Comma separated value (CSV) (Module 2..4)This format is an unstandardised plain text format for tabular data with the following properties:
- each value delimited (separated) by a comma " , " or other special characters
- the first line may contain the variable names
-each line contains a single record
- the position of values in the line corresponds to the variable - 49JavaScript Object Notation (JSON) (Module 2.4)This format is becoming one of the most popular ways to transfer data, particularly on the web. If you retrieve a dataset from an online provider, it may be in JSON format. The JSON format allows for more complicated and detailed datasets, with a nested or hierarchical structure.
- 50Data dictionary (Module 2.4)A data dictionary is a document that provides:
- a list of variables
- the data type for each variable
- set of possible values, if categorical
- a written description or explanation of each variable - 51KDDA (Module 3.1)Knowledge discovery via data analytics (KDDA) is understood as an entire process to identify valid, non-trivial, previously unknown, interesting relationships/patterns that exists in large quantities of data.
- 52CRISP-DM (Module 3.1)Cross Industry Standard Procedure for Data Mining (CRISP-DM)
- the most commonly used data mining process model
-guides the user through each step of data mining leading to knowledge discovery
- 53Phase 1: Business understanding CRISP-DM (Module 3.1)The objective of the business understanding phase is to clearly define the business problem or, the business opportunity presented through knowledge discovery.
Other objectives of this phase include - determining the success criteria associated with the knowledge discovery exercise. It includes
Consequently, the main outcome of this phase is a clear statement of the problem or the opportunity that knowledge discovery presents to the business. For example, "increase online sales to existing customers by 7%."
The problem statement should be part of a project plan, along with a summary of resources - people, skills, capital, and technologies - required to conduct the knowledge discovery process. A timeline should also be included in the project plan. - 54Phase 2: Data understanding CRISP-DM (Module 3.1)The objective of this phase is to determine whether the organization has the right kind of data to undertake the knowledge discovery exercise. Assuming that relevant data is available, this phase will involve describing the data; articulating the features available in the data; understanding the format of the data; determining the method to deal with data depending on whether it is small data stored in excel files or big data that may be real-time and streaming.
Exploratory data analysis (EDA) is carried out, Data quality is assessed, and Metadata ("data about data") is looked at in this phase. - 55Phase 3: Data preparation CRISP-DM (Module 3.1)Once we understand the data, the next step for us is to finalize the dataset we need for the knowledge discovery exercise. This step includes determining the selection criteria to choose the features we want to use from the data. Selecting appropriate features from the data is dependent on a variety of factors such as the problem objectives, goals, data quality, and other constraints such as the 4 V’s.
Data preparation is also where we clean and fix data problems of quality associated with the data, may even have to construct new data, discretize data or impute data - 56Phase 4: Modeling CRISP-DM (Module 3.1)- Select most appropriate modeling technique (decision trees, clustering algorithms, classification models, etc.)
- Develop a test design: partition data into training, test, and validation datasets
- Assess the models: evaluate against objective and success criteria,, look at model quality, and decide which model is best - 57Phase 5: Evaluation CRISP-DM (Module 3.1)- Evaluate the results of the modeling phase
- Assess the degree to which the models meet the stated business objectives
- Review the process used to build the models
- Quality Assurance (QA); check all features used in modeling are available to operationalize, no inherent biases, modeling exercise is repeatable
- Decision to finish project and deploy or revisit earlier steps - 58Phase 6: Deployment CRISP-DM (Module 3.1)- Plan the deployment of the models in the production environment
- Organization can start to use the models for knowledge discovery
- Monitor the performance of the models while they are in production
- Wrap up the project by creating a project report and reviewing the project outcomes based on feedback from the end users - 59Snail Shell vs. CRISP-DM (Module 3.1)Snail Shell includes:
- Problem Formulation
- Maintenance - 60EDA - Notation (Module 3.2)
- 61EDA - Terminology (Module 3.2)
- 62Goal of EDA (Behren) (Module 3.2)
- 63EDA techniques
- 64Data quality - Missing Completely at Random (MCAR) (Module 3.3)Missing-ness is independent of the observed and unobserved variables.
Ex. Any two individuals, regardless of their values of disease status, level of exposure, and age, have the same probability of having a missing value for exposure.
Strongest assumptions -> Easiest to model - 65Data quality - Missing at Random (MAR) (Module 3.3)Missing-ness depends on observed variables.
Ex. Any two individuals with the same disease status and age have the same chance of having a missing exposure value, regardless of how large or small their actual level of exposure. - 66Data quality - Missing Not at Random (MNAR) (Module 3.3)Missing-ness depends on unobserved variables.
Ex., Even among individuals with the same disease status and age, the chance of having a missing exposure value depends on their level of exposure.
Weakest assumptions -> Hardest to model - 67Five-number summary (Module 3.4)In a univariate setting, the distribution of a numerical variable can be summarised using five numbers:
- minimum (smallest value in the sample)
- first quartile
- median
- third quartile
- maximum (largest value in the sample) - 68Boxplot (Module 3.4)A box-plot (or box-and-whisker-plot) is a graphical summary of the distribution of a numerical variable based on order statistics.
- 69Sample percentile (Module 3.4)The k-th sample percentile is another name for the k% sample quantile.
- 70Outliers (Module 3.4)Outliers are observations that deviate significantly from the systematic characteristics or patterns of the sample. There is not a unified definition of when an observation should be considered an outlier. Any filter for outliers must be application dependent.
Information
- Description
- Business Analyst BUSS6002 Flashcards on Mid-Semester Exam, created by Triton Pelstring on 09/09/2023.
- Created by
- Triton Pelstring
- Views
- 1
- Rating