RE: Altair slc supplementing rapid miner with genetic modeling using r and python
Too long to post on a list, see github
https://github.com/rogerjdeangelis/utl-altair-slc-supplementing-rapid-miner-with-genetic-modeling-using-r-and-python
graphics
https://github.com/rogerjdeangelis/utl-altair-slc-supplementing-rapid-miner-with-genetic-modeling-using-r-ga-package/blob/main/plot1_timeseries_merge.pdf
Rapid Miner Post
https://community.altair.com/discussion/44504/rapid-miner?tab=all
TWO SOLUTIONS
- r package GA
- python
Note: This is a very simple example, close to fitting an AR(2) model.
But GA(genetic algorithm) becomes more useful when: (none demonstrated here)
1 Constraints: Force parameters to sum to 1 (for integrated processes)
2 Non-linear objectives: Minimize VaR instead of MSE
3 Multiple objectives: Balance forecast accuracy with parameter parsimony
4 Complex structures: AR with neural network components
Problem (what you can expect from this example of genetic algorithm iterative processing)
Stock Trading
Strong lag(1) price momentum: 0.61(lag1) coefficient means 61% of today's return persists tomorrow
Moderate lag(2) price momentum 0.32(lag2) 32% of day 1 return persists
Mean reversion: -0.04 at lag 3 suggests slight reversal after 3 days
Trading implication: Momentum strategy might work for 1-2 days
Weather
If its sunny today there is a 61% chance it will be sunny tomorrow
If its sunny today there is a 31% chance it will be sunny the day after tomorrow
Possible cloudy in two days
