Lab I : Probabilistic Modeling, Inference and Decision Making with Netica
Objectives
Netica is an application software for building Bayesian networks and influence diagrams, and for performing probabilistic inference and decision making. It also supports some limited capabilities in Bayesian network learning.
Learning Outcomes
At the end of the lab session, you will be able to:
- Build a simple Bayesian Network using the GUI and input the relevant data.
- Compile the network and perform various types of probabilistic inference and experiment with the network.
- Observe conditional independence and dependence through experimentation.
- Perform sensitivity analysis on the network using mutual information.
- Build a simple influence diagram by extending the Bayesian network you have built.
- Find the optimal decision policy from the influence diagram.
- Perform probability learning from data
Lab Schedule
You may do this on-line lab at your own time after completing the lecture videos on Section 5.1 (Bayesian Networks)
Lab Instructions and Materials
Lab II: Decision Modeling and Analysis using DPL
Objectives
DPL is an industry standard software for performing professional decision analysis by consultants. It supports decision modeling using both influence diagrams and decision trees. It can model asymmetric decision models, has extensive tools for generating risk profiles, performing sensitivity analyses (Tornado diagrams, Rainbow Diagrams, etc), expected value of information & control analyses, real options (or managerial flexibility) valuation analysis, etc. It can be linked to Excel spreadsheets and also has a powerful programming language. You can create a decision model by just writing codes with a text editor.
Learning Outcomes
At the end of this lab, you will be you will be able to:
- Build decision models using the graphical model editor and input the relevant data such as probabilities and utility function.
- Solve the decision models to find optimal decision policies
- Generate and interpret risk profiles
- Perform and interpret expected value of information and control on uncertain variables.
- Perform one-way sensitivity analysis and generate rainbow diagrams.
- Build sequential decision models with asymmetric structure.
- Reorder and modify decision tree structure generated by DPL.
- Perform expected value of imperfect information analysis.
- Perform two-way sensitivity analysis and generate two-way rainbow diagrams.
Lab Schedule
You may do this on-line lab at your own time after completing Chapter 5.