Advanced Analytics, Machine Learning, and AI

Introduction to Analytics 2022-2023 Lesson 4

• This lesson covers advanced analytics, machine learning, and AI.

Learning Objectives

• Understand the meaning of prediction.
• Recognize predictor and response variables.
• Discuss data used for advanced analytics.
• Distinguish types of predictive models.
• Differentiate between machine learning (ML) and artificial intelligence (AI).
• Explain the two types of AI.
• Learn to ask advanced analytics questions.

Agenda

• Topics covered in this lesson:
• What is prediction?
• From prediction to decision
• Predictor and response variables
• Data used in advanced analytics
• Types of predictive models
• Advanced analytics questions
• ML and AI; two types of AI
• Group assignment introduction

Types of Advanced Analytics

• Descriptive Analytics: Examines data to answer "What happened?" or "What is happening?"
• Diagnostic Analytics: Examines data to answer "Why did it happen?"
• Predictive Analytics: Examines data to answer "What is going to happen?" or "What is likely to happen?"
• Prescriptive Analytics: Examines data to answer "What should be done?" or "What can we do to make X happen?"

Prediction Defined

• Prediction: Using available information to generate information you don't have.
• Predict: Foretell based on observation, experience, or a scientific reason.
• In science, prediction is a rigorous and quantitative statement forecasting what will happen under specific conditions.

Predictive Modeling

• Used to solve business problems by:
• Improving efficiency
• Making better decisions
• Enabling new organizational capabilities

Predictive Modeling: Reframing Business Problems

• Predictive modeling helps reframe these business problems:
• Determining what someone is likely to do in a given scenario.
• Determining what is likely to happen if certain conditions are met.
• Estimating how much is expected and when.

Value of Prediction

• Prediction assists to solve a business problem.
• Analytics involves examining information to uncover knowledge that helps business people make informed decisions.

Predictive Modeling: Examples

• Predictive modeling applications:
• Pipe breakdown prediction.
• Survival analysis, which predicts occurrence and timing of events like customer churn or loan defaults
• Network analysis, which targets individuals in a network for advertising
• Net lift response modeling, which predicts the likelihood of customers responding to promotions

Prediction vs. Decision

• Decisions can be:
• Automated based on model scores
• Partially automated for standard behaviors, with rules for exceptional cases
• Not automated, requiring human judgment using model scores and predictions

Human Role in Analytics

• Humans make decisions based on analytics.
• Humans implement automation as is relevant to the analytics data.

Analytics Maturity

• Maturity will be determined by:
• the degree of decision process automation.
• the level of human involvement in decision-making.

Predictive Modeling and Variables

• Model: A mathematical representation of an object or process.
• Predictive modeling: Uses mathematical and computational methods to develop predictive models.
• Predictive modeling examines datasets for underlying patterns and calculates outcome probabilities.
• It focuses on understanding the relationships between predictor data and behavioral (response) data.
• Predictive analytics process: Analyzes data to identify how predictor data differentiates behaviors and predicts outcomes.

Predictor and Response Variables

• Predictor variable: The variable used to predict the response.
• Predictor variable is also known as: independent variable, feature, attribute, or X-variable.
• Response variable: The variable the model aims to predict.
• Response variable is also known as: dependent variable, outcome, target, or Y-variable.

Sources of Predictor Variables

• Existing metadata provides documentation describing business data, relationships, and constraints.
• The current decision-making process provides insight into what data is used, and what parameters/attributes are included in reports.
• Subject matter experts working in operational areas understand what information is important and relevant.
• External data sources, such as credit rating agencies or business associations, have knowledge of relevant data.
• Research and peer experience can show what data is used by competitors for similar problems.

Data Used for Advanced Analytics

• Primary behaviors: Information about past behavior of the same type is used to predict the behavior you want to predict. Example: Predicting burglary if someone has committed a burglary before.
• Secondary behaviors: Information about past behaviors that are similar (but different) to the behavior you want to predict. Example: Minor crimes and offenses.
• Tertiary behaviors: Information about past behaviors that do not have an obvious connection to the behavior you want to predict. Example: Selling items on eBay.
• Geo-demographic: Information about a person's state of being. Example: age, income, occupation, appearance, education level.
• Associate data: Information about a person's connections, including their geo-demographic data and behaviors. Example: Partner or friends that committed a burglary.
• Sentiments: A person's attitudes, feelings, and opinions. Example: Social media likes/dislikes and blog posts.
• Network: Information about the nature of connections between a person and their associates. Example: Majority of friends have committed criminal offenses.

Machine Learning (ML)

• Supervised learning: Predicts outcomes using labeled data, learning from the data to detect patterns.
• A "training set" is available, offering correct answers or "labels" for given scenarios or "input variables."
• Unsupervised learning: Predicts outcomes using unlabeled data.
• Aims to model underlying data structure or distribution in order to learn more about the data, where the correct answer is unknown

Types of Machine Learning Models

• Supervised learning models:
  • Classification models: Predict the probability of an event via estimate probability.
  • Regression models: Predict a quantity via estimate value.
• Unsupervised learning models:
  • Clustering: Finds a structure or pattern in a collection of uncategorized data.
  • Association: Discovers relationships between variables.

Predictive Models: Supervised Learning

• Classification problem: Classifies an instance.
  • Description: Predicts the probability of an event from a discrete list of events/possibilities.
  • Output: Probability in %.
  • Response variable: Categorical.
  • Example: Logistic regression (binary problems), decision tree, random forest, Naive Bayes, SVM, etc.
• Regression problem: Finds a function that fits the data with the least error.
  • Description: Predicts a quantity or amount.
  • Output: estimate (numeric value)
  • Response variable: Numerical.
  • Example: Linear regression, polynomial regression, ridge regression, etc.

Supervised vs. Unsupervised Machine Learning

• Supervised learning: Algorithms trained using labeled data. A simpler, more accurate and trustworthy method.
• Unsupervised learning: Algorithms used against data that is not labeled. Computationally complex, less accurate and trustworthy method.

Predictive analytics questions

• What is someone likely to do in a given scenario?
• What is likely to happen if...?
• How much....is expected when...?
• What is the expected for...?
• What is the likelihood that someone will choose... if...?
• What is the probability that... belongs to... category?

Machine Learning vs. AI

• Artificial intelligence (AI) is the theory and development of computer systems able to perform tasks normally requiring human intelligence.
• AI includes visual perception, speech recognition, decision-making, and translation between languages.
• Machine learning (ML) is an AI discipline that allows computers to handle new situations through analysis, self-training, observation, and experience.
• Deep learning (DL) is a subfield of machine learning concerned with algorithms inspired by artificial neural networks.

Types of AI

• Narrow AI: Intended for a narrowly defined or single task. All AI that exists today is narrow.
• General AI: Intended to be capable of a broad set of tasks comparable to what can be done by a human.
• Narrow AI solves problems framed as prediction problems.
• General AI intends to solve problems that a human has the capacity to solve.

Applications of Narrow AI

• Analytics applications examples:
• Recommender systems based on search history
• Targeted advertising on behavior
• Speech recognition
• Autocomplete
• Gaming
• Fraud and risk detection
• Logistics
• Sports cameras

Module 7 of the Program

• Module 7 covers:
• Managing analytics projects
• Roles in analytics projects
• Quantitative Research Methods Course:
• Descriptive statistics
• Probability and distributions, hypothesis testing
• Big Data Course:
• Loading, storing, and processing big data
• Querying big data
• Big data manipulation, analysis, and visualization
• Data Analytic Tools course:
• Traditional and advanced analytics tools
• Data manipulation
• Quantitative analysis and statistical techniques