Execution & Results - Heisenberg0203/GETS GitHub Wiki

Executing GETS (Please make sure that you are in src folder)

python ponyge.py --parameters path_to_parameter_file.

Command to run different Time series forecasting model is given below. By default gets execute with parameters as given in Parameters File.

Note: You may also experiment with different parameters and storing the text file in the parameters folder.

  1. Simple Exponential Smoothing

python ponyge.py --parameters gets/ses.txt

  1. Holts Exponential Smoothing

python ponyge.py --parameters gets/ses.txt

  1. Holts Winter Exponential Smoothing

python ponyge.py --parameters gets/tes.txt

  1. Moving Average

python ponyge.py --parameters gets/ma.txt

Executing GETS N times.

Due to Stochastic Nature of Grammatical Evolution, the program may end up in local optima for a single run. Multiple runs of GE are always preferable.

Command to execute GETS n time:

FOR /L %i IN (1,1,N) DO python ponyge.py --parameters regression.txt

Replace N with the number of runs you want to execute GETS. ( 30 is preferable)

RESULTS:

The results for the GETS will be available in the results folder. The directory of results contains the following files:

  1. avg_fitness.pdf: The graph of Average_Fitness vs Generation

  2. avg_fitness_array.txt: Average Fitness values for each generation

  3. best.txt: It contains the Phenotype (Code) generated along with all the parameters required to make time-series forecast. It also contains Genotype( Sequence of Codons) used by GETS to generate above phenotype along with Training and Test fitness.

  4. best.pdf: The graph of Best_fitness vs Generations

  5. best_fitness_array.txt: Best Fitness values for each generation

  6. parameters.txt: Parameters used by Genotype to generate Phenotype.

  7. stats.tsv: Complete stats of each generation.