I am new to JGAP. I have searched the documentation and archived lists for an answer to this question. Please forgive me if I have missed something.
I am building a marketing/purchasing simulation. It attempts to simulate the motivations for consumers to select and purchase products based on their preferences (for product features, price, availability, etc) - in response to various marketing events (launches, promotions, mailings, advertisements, etc). The simulation will run against real market data collected over the previous 2 years. The overall goal is to optimise marketing spend in response to customer preferences. However, within this simulation, the 'goal' of the consumer is to satisfy as many preferential criteria as possible as they are exposed to the various marketing events.
During a marketing event, each virtual-consumer is presented with a product promotion that potentially satisfies various consumer criteria. Each of the consumer's preferences are encoded as follows (e.g. for Price):
A) Purchase-Threshold value (double) - a value that must be satisfied for a purchase to be made
B) Range to be applied to A (double) - for example, if value A is 75 and B is 10, then a purchase will be made if the price is between 75 +/- 10 i.e. between 65 and 85
C) Weighting (Integer) to be applied pro-rata to the preference before adding it to other preferences to achieve an overall 'motivation' value. Using the above example, Price may be unimportant compared to other consumer preferences and so be given a low comparitive weighting.
A virtual consumer has 12 preferences, each of which consists for the values a, b and c as described above.
My question is this: should I have 36 separate genes (representing the 3 x 12 values), or should (a, b, c) be combined in a Composite gene, leading to 12 Composite genes per consumer?
I have read the documentation for Composite genes and I understand how to use them. What is not clear is whether the genetic operators apply differently to composite genes as opposed to ordinary genes. What is the advantage of using Composite genes?
Many thanks to the JGAP contributors for a fantastic GA environment. I hope someone can help with my question.