Business Intelligence
wide variety of tools, applications, and methodologies to run queries, create report, dashboards, and data visualizations for decision makers; to understand the business and determine relationships among internal and external factors "BI answers what happened", studies historical data
Business Analytics
software tools and applications used to build models to create scenarios, understand current events, and predict future state "BA answers why it happened and whether it will happen again", looking forward to understand why
Benefits of BA
improve operational efficiency, better understand customers, project future outcomes, gain insights, measure performance, discover hidden trends, etc
Cycle of Analytics
data access, discovery, exploration, and information sharing to react to change questions and expectations
Traditional BI v.s. Modern BI
traditional BI was a top-down approach that led to slow, frustrating reporting cycles, modern allows multiple levels of users to customize dashboards and create reports
Data Mining
process of discovering meaningful correlations, patterns, and trends by sifting through large amounts of data; not obvious relationships, discover patterns, predictive [OLAP]
What can data mining do and not do?
can help you find patterns and relationships (cannot eliminate the need to understand your data and know your business), can discover hidden information and data (cannot tell you the value of that information)
Data Warehousing
proper data cleansing and preparation can be facilitated by a data warehouse, applicability to problem
Descriptive Data
what happened? hindsight
Predictive Data
what will happen? insight
Prescriptive Data
how can we make it happen? foresight
Diagnostic Analytics
why is it happening?
Business Intelligence Criteria
Accuracy (of inputs and outputs), valuable insights, timeliness (of data going in and insights coming out), actionable
Business Analytics Methods
clustering (recognizing direct groupings or subcategories within the data), classifying, estimating and predicting, and affinity grouping (special kind of clustering that identifies events that occur simultaneously)
Best Practices with BA
know the objective, define criteria for success and failure, select methodology, and validate models using predefined criteria
Challenges with BA
risk of spending money and time chasing poorly defined problems, mistake noise for true insight, not accessing correct data, time and effort to clean data, etc
Data Mining Process
problem definition, data gathering & preparation, model building & evaluation, knowledge deployment (insight, scoring, extraction of model details, integration of tools)
Artificial Intelligence
the ability of a machine to perform cognitive functions we associate with human minds, such as reasoning, learning, and problem solving
Norvig and Russel Four Types of Approaches (defining AI)
thinking humanly, thinking rationally (thought processes and reasoning), acting humanly, acting rationally (behavior)
Four Types of AI
Reactive machines (Deep Blue (chess)), limited memory (reinforcement, RNN, E-GAN, transformers), theory of mind (how others feel), self-awareness (understand existence)
General AI
"strong AI" still much theoretical, e.g. GPT-3 is closest
Narrow AI
limited context and performs single task extremely well; machine learning and deep learning (self-driving cars)
Machine Learning (supervised and unsupervised)
algorithms that detect patterns and learn how to make predictions, recommendations, and decisions
Deep Learning
type of machine learning, a biologically inspired neural network architecture; data comes pre-labeled
Superintelligence
surpasses humans in every way, hypothetical situation
History of AI
First look in 1940s, (1950) Alan Turing, 2010s had big gains, 2020s is GPT-3
Pros of AI
Improves productivity and efficiency, while reducing potential for human error
Cons of AI
Development costs and possibility for automation to replace human jobs
AI Today
uses predictive analysis, knowledge creation, customer insight, helps with pricing, supply chain, R&D, credit card fraud, healthcare, and Netflix recommendations
Concerns Regarding AI
(1) adverse impact of AI on labor (2) biases (3) lethal autonomous weapon systems (4) superintelligence
AI & Crime
AI can detect gunfire, predict crime spots and even who will commit a crime; however, criminals also take advantange of AI
EU Guidelines & AI
the European Commission published seven principles to create "trustworthy" AI programs as the baseline for companies; formulates best practices and advance public understanding of AI
Algorithmic Bias
gender bias, racial bias, age discrimination, and other human characteristics; prejudice based on a particular categorical distinction
AI systems learning
Learn based on training data, which may have skewed human decisions or represent historical or social inequities
"Faces in the Wild"
Facial recognition software considered the benchmark for testing facial recognition, however there is imbalance since the data is has was 70% male and 80% white