Chapter 19 Summary
Analyzing Decisions Using Datasets with Multiple Attributes: A Machine Learning Approach
Janusz Wojtusiak - George Mason University
Farrokh Alemi - Georgetown University
This chapter is too old. When it speaks of ‘machine learning’ it actually is referencing ‘rule based models’. Specifically AQ21. Their most ‘uptodate’ algorithm mentioned is from 2006. That is 12 years ago.
Much has changed since this chapter was written.
Decision Analysis in Healthcare
The chapter proposes the following steps as a theoretical framework of decision analysis:
1. Define goal: who is the person or group making de decision. What is the decision being addressed?
2. Define alternatives: what are all the possible ways the goal can be reached
3. Define perspective of the analysis: an analysis can occur from the perspective of different stakeholders (the pharmaceutical company, the patient, the doctor). Approaching the goal from each of these groups will produce a different analytical framework and results. If desired it may be possible to join the different viewpoints in a “criteria decision analysis (MCDA)”.
4. Select data sources: are there databases? expert opinion? or unknowns?
5. Build models of alternatives: we now have the data, and alternatives that can reach the goal, now different methods can be applied to try and build these models
6. Use models to make predictions/evaluate alternatives:
7. Perform sensitivity analysis of the obtained results:
8. Make a recommendation: