probabilistic_hybrid_optimizer.py

from recommenders.hybrid.Probabilistic_Hybrid import Probabilistic_Hybrid
from recommenders.hybrid.borda_hybrid import Borda_Hybrid
from skopt import gp_minimize
from skopt.space import Real, Integer, Categorical
from skopt.utils import use_named_args
import data

param_list = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30]

"""

Optimizer intended for the Probabilistic_Hybrid class.

"""

@use_named_args([
    Categorical(param_list, name='last_interaction'),
    Categorical(param_list, name='catboost_066693_local'),
    Categorical(param_list, name='location_subm'),
    Categorical(param_list, name='min_price_based'),
    Categorical(param_list, name='xgb14f0665local')
])
def objective(last_interaction, catboost_066693_local, location_subm, min_price_based, xgb14f0665local):
    params = {
        'last_interaction': last_interaction,
        'catboost_066693_local': catboost_066693_local,
        'location_subm': location_subm,
        'min_price_based': min_price_based,
        'xgb14f0665local': xgb14f0665local
    }

    model = Probabilistic_Hybrid(params, mode='local')
    model.fit()
    sub = model.recommend_batch()
    MRR = model.score_sub(sub)

    print(f'Iteration parameters: '
          f" - last_interaction= {last_interaction} - catboost_066693_local= {catboost_066693_local} - location_subm= {location_subm} - min_price_based= {min_price_based} - xgb14f0665local= {xgb14f0665local}")

    return -MRR


space = [
    Categorical(param_list, name='last_interaction'),
    Categorical(param_list, name='catboost_066693_local'),
    Categorical(param_list, name='location_subm'),
    Categorical(param_list, name='min_price_based'),
    Categorical(param_list, name='xgb14f0665local')
]

res_gp = gp_minimize(objective, space, n_calls=250, n_random_starts=10, random_state=17, verbose=True)

print("""Best parameters:
- last_interaction= %d
- catboost_066693_local= %d
- location_subm= %d
- min_price_based= %d
- xgb14f0665local= %d
"""% (res_gp.x[0], res_gp.x[1], res_gp.x[2], res_gp.x[3], res_gp.x[4]
      ))