Tutorial
This is a quick tutorial based on examples to learn to use quickly the differents algorithms of the library
Genetic
Genetic is a metaheuristic algorithm wich mean that it's not deterministic, so every time the algorithm output will be a different solution near of the unknown best solution.
This Algorithm is focused in solve the MTSPTW problem (Multi Travel Sales Problem with Time Windows). This problem is about a person that need to cover a number of points to visit but have a limit of working hours (time constraint) and extra cost. If it's the case, this algorithm is the perfect solution:
Load the modules
First of all we load the modules that we will use:
from sklearn_route.datasets import load_barcelona
from sklearn_route.preprocessing import matrix_to_dict
from sklearn_route.genetics import Genetic
If you want to know more about these modules you can visit:
Pre-processing data
Here we put the data in the format required by the algorithm. In this case we use the matrix_to_dict function to transform our id_points cost in fuel and in time to dicts of dicts.
You can see the docs of all functions, using the help() command. Also there are more detailed examples in the docs webpage.
df_barcelona = load_barcelona()
#Dataset - id of origin - id of destiny - column to transform (in this case the hour)
time_matrix = matrix_to_dict(df_barcelona, "id_origin", "id_destinity", "hora")
#Dataset - id of origin - id of destiny - column to transform (in this case the cost)
cost_matrix = matrix_to_dict(df_barcelona, "id_origin", "id_destinity", "hora")
#Create a random route, it's will needed to initiate the algorithm
route_example = list(dict.fromkeys(cost_matrix_df).keys())
Running the Algorithm
Here we see how the algorithm operation is the same as Scikit Learn. You can see all the Hyperparamenters, what they do, and examples with the builtin help() function. Also there are more detailed examples in the docs webpage.
In this case we choose an algorithm with Methaeuristic aproach to the problem. There are other algorithms in the package that can solve this problem or have another approach.
#Instantiate the algorithm
ga = Genetic(p_m = 0.3, pop=400, gen=2000, k=5, p_c early_stoping=100, max_time_work=6, extra_cost=12.83)
#random route - time_matrix - cost_matrix
result = ga.fit(route_example, time_matrix, cost_matrix)
#Printing the best route
print(result)
#Printing the loss function
print(ga.history_)
In the Genetic Algorithm we have a lot of Hyperparameters to config, for a full understanding of the Algorithm please refer to Genetics