def __training(self, extractor: ExamDropExtractor) -> LogisticRegression:
""" Execute the training phase of the Exam Drop Layer.
Arguments:
extractor (ExamDropExtractor): The extractor which delivers the training data.
Returns:
The trained machine learning model used to make predictions.
"""
train_input, train_output, train_weights = extractor.get_train_data()
classifier = LogisticRegression(solver="lbfgs")
classifier.fit(train_input, train_output, train_weights)
return classifier
def predict(self, predict_season: int, latest_episode: int,
train_seasons: Set[int]) -> Dict[Player, MultiLayerResult]:
available_seasons = EXAM_DATA.keys()
train_seasons = train_seasons.intersection(available_seasons)
if predict_season not in available_seasons:
return EmptyMultiLayer().predict(predict_season, latest_episode,
train_seasons)
extractor = ExamDropExtractor(predict_season, latest_episode,
train_seasons,
self.__anova_f_significance,
self.__pca_explain, self.__max_splits)
classifier = self.__training(extractor)
predict_data = extractor.get_predict_data()
if not predict_data:
alive_players = EXAM_DATA[predict_season].get_alive_players(
latest_episode)
return EmptyMultiLayer(alive_players).predict(
predict_season, latest_episode, train_seasons)
return self.__predict(predict_season, latest_episode, predict_data,
classifier)
def predict(self, predict_season: int, latest_episode: int,
train_seasons: Set[int]) -> Dict[Player, MultiLayerResult]:
available_seasons = EXAM_DATA.keys()
train_seasons = train_seasons.intersection(available_seasons)
if predict_season not in available_seasons:
return EmptyMultiLayer().predict(predict_season, latest_episode,
train_seasons)
extractor = ExamDropExtractor(predict_season, latest_episode,
train_seasons, self.__min_cluster_size,
self.__spline_curves,
self.__random_generator)
in_classifier, out_classifier = self.__training(extractor)
predict_data = extractor.get_predict_data()
if not predict_data:
alive_players = EXAM_DATA[predict_season].get_alive_players(
latest_episode)
return EmptyMultiLayer(alive_players).predict(
predict_season, latest_episode, train_seasons)
return self.__predict(predict_season, latest_episode, predict_data,
in_classifier, out_classifier)
def __training(
self, extractor: ExamDropExtractor
) -> Tuple[LogisticRegression, LogisticRegression]:
""" Execute the training phase of the Exam Drop Layer.
Arguments:
extractor (ExamDropExtractor): The extractor which delivers the training data.
Returns:
Two trained machine learning model used to make predictions. The first model makes predictions for cases
where a player is in the answer and the second model makes predictions for cases where a player is out the
answer.
"""
train_input, train_output, in_answer, train_weights = extractor.get_train_data(
)
in_input = [
ti for ti, answer in zip(train_input, in_answer) if answer == 1.0
]
in_output = [
to for to, answer in zip(train_output, in_answer) if answer == 1.0
]
in_weights = [
tw for tw, answer in zip(train_weights, in_answer) if answer == 1.0
]
in_classifier = LogisticRegression(
solver="lbfgs",
max_iter=self.MAX_TRAIN_ITERATIONS,
random_state=self.__random_generator)
in_classifier.fit(in_input, in_output, in_weights)
out_input = [
ti for ti, answer in zip(train_input, in_answer) if answer == 0.0
]
out_output = [
to for to, answer in zip(train_output, in_answer) if answer == 0.0
]
out_weights = [
tw for tw, answer in zip(train_weights, in_answer) if answer == 0.0
]
out_classifier = LogisticRegression(
solver="lbfgs",
max_iter=self.MAX_TRAIN_ITERATIONS,
random_state=self.__random_generator)
out_classifier.fit(out_input, out_output, out_weights)
return in_classifier, out_classifier
from Data.PlayerData import get_is_mol
from Layers.ExamDrop.ExamDropEncoder import ExamDropEncoder
from Layers.ExamDrop.ExamDropExtractor import ExamDropExtractor
from Tools.Encoders.NaturalSplineEncoding import NaturalSplineEncoding
from Tools.Encoders.StableDiscretizer import StableDiscretizer
from sklearn.linear_model import LinearRegression
import matplotlib.pyplot as plt
import numpy as np
import sys
TRAIN_SEASONS = {5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22}
NUM_CURVES = 4
train_data = []
for season in TRAIN_SEASONS:
train_data.extend(ExamDropExtractor.get_season_data(season, sys.maxsize, True))
train_input = np.array([ExamDropEncoder.extract_features(sample, sys.maxsize) for sample in train_data])
train_output = np.array([1.0 if get_is_mol(sample.selected_player) else 0.0 for sample in train_data])
m = ExamDropEncoder.NUM_CONTINUOUS_FEATURES
for column, feature_name in zip(train_input.T[-m:], ExamDropEncoder.FEATURE_NAMES[-m:]):
trans_data = column[:,np.newaxis]
spline_encoder = NaturalSplineEncoding([NUM_CURVES])
trans_data = spline_encoder.fit_transform(trans_data)
in_answer_input = [row for row, data in zip(trans_data, train_data) if data.selected_player in data.answer]
in_answer_output = [to for to, data in zip(train_output, train_data) if data.selected_player in data.answer]
regression = LinearRegression()
regression.fit(in_answer_input, in_answer_output)
X = np.array([pi for pi in sorted(set(column))])