What types of analytics are there?
Prescriptive
-> plan under uncertainty, using robust optimization, heuristics
Predictive
-> evaluate scenarios and strategies, using forecasts, classification, regression, inductive statistics, simulation
Descriptive
-> exploratory and descriptive data analysis, using data warehouse, OLAP, clustering, statistics, data visualization
Descriptive analysis
- prepares and analyzes historical data
- identifies patterns from samples for reporting of trends
Descriptive analysis
Methods
- Descriptive statistics, e.g. moments
- Data mining, e.g. clustering
- Visualization
Descriptive Analysis
Example
- Compare mean distributions and variation to observe systematic differences between forecasts and sales
Predictive Analysis
- Predicts future probabilities and trends
- Finds relationships in data that may not be readily apparent with descriptive analysis
Predictive Analysis
Methods
- Inductive statistics, e.g. exponential smoothing and regression
- Data mining for predictions, e.g. on customer responses
Prescriptive Analysis
- Evaluates and determines new ways to operate
- Targets business objectives given a set of contraints
Prescriptive Analysis
Methods
- Evaluates and determines new ways to - Mathematical modeling
- Analytical optimization
- Heuristics and meta-heuristics
Prescriptive Analysis
Example
Multi-criteria optimization to allocate the assignment of “Blindbookings” to specific flights
Data Analysis in Advanced Analytics
Descriptive:
- statistics to describe the data
- visualizations
Model driven
- also: inductive
- pick the “best” model for your data, parametrize it, evaluate the results
Data driven
- also: exploratory, data mining
- consider patterns in the data using computational algorithms
-> for maximum impact, combine model driven and data driven approaches
Simulation in Advanced Analytics
Components
Status Quo
Basic Scenario
What-if Scenario
Simulation in Advanced Analytics
In the real world there’s a status quo: the system as it currently functions in the real world.
Based on this you calibrate a basic scenario for a simulation model. In the basic scenario models describe the structure and data analysis fills parameters. Models need validation, though.
The what-if Scenario is derived from the basic scenario through relative changes in parameters or structural changes in algorithms. The what-if scenario can be interpreted in its relationship to the real future.
Optimization in Advanced Analytics
What kinds of optimization are there?
deterministic
robust
Optimization in Advanced Analytics
Deterministic optimization
Deterministic (not robust) optimization determines the best solution for a single parameter set
-> this is great when you know exactly what the parameters are going to be
Optimization in Advanced Analytics
Robust optimization
Robust optimization considers a number of possible parameter sets (scenarios) and determines the solution that is the best “compromise” for all
- > this is great, when you have an idea of possible scenarios
- > even better if you can estimate the probability per scenario
Teams for analytics
Success factors of Teams
- Small groups
- Complementary abilities
- Everyone has a clear view of the joint objective and is motivated to achieve it
- Everyone puts in the same amount of work using similar tools
- Everyone is bound to hold themselves and the others to accountable
-
1 - Introduction16
-
2 - Forecasting and Estimation37
-
3 - Data Warehouses and OLAP44
-
4 - Unsupervised learning: clustering and association rules37
-
5 - Supervised Learning, Classification26
-
6 - Supervised learning, classification13
-
7 - Validating and Evaluating Data Models19
-
8 - Data Visualization26
-
9 - Simulation Modelling45
-
10 - Stochastic Simulation Modelling19
-
11 - Validating simulation models28
