def main():
print("Setting up dataset and evaluator...")
# Setup dataset and evaluator
dataset = Dataset(load_tags=True,
filter_tag=False,
weight_tag=False)
dataset.set_track_attr_weights_2(1.0, 1.0, 0.0, 0.0, 0.0,
1.0, 1.0, 0.0, 0.0)
ev = Evaluator(seed=False)
ev.cross_validation(5, dataset.train_final.copy())
for i in range(0, 5):
print("Training models for fold {}-th".format(i))
train_models(dataset, ev, n_clusters=20)
def main(n_evaluations):
print("Setting up dataset...")
# Setup dataset and evaluator
dataset = Dataset(load_tags=True,
filter_tag=False,
weight_tag=False)
dataset.set_track_attr_weights_2(1.0, 1.0, 0.0, 0.0, 0.0,
1.0, 1.0, 0.0, 0.0)
# Run mAP evaluation a number of times to reduce the variance
# of the results
n_clusters = 20
best = np.zeros((4, n_clusters), dtype=int)
for i in range(n_evaluations):
print("Running cross validation no. {}".format(i))
print("Setting up the Evaluator...")
ev = Evaluator(seed=False)
ev.cross_validation(5, dataset.train_final.copy())
folding_best = train_models(dataset, ev,
n_folds=5, n_clusters=n_clusters)
best += folding_best
top = top_over_cluster(best)
_print_top(top)
def evaluateMap():
from src.utils.evaluator import Evaluator
dataset = Dataset(load_tags=True, filter_tag=False, weight_tag=False)
dataset.set_track_attr_weights_2(2.0, 2.0, 0.0, 0.0, 0.0, 2.0, 2.0, 0.0,
0.0)
# seed = 0xcafebabe
# print("Evaluating with initial seed: {}".format(seed))
ev = Evaluator(seed=False)
ev.cross_validation(5, dataset.train_final.copy())
cbf = ContentBasedFiltering()
for i in range(0, 5):
urm, tg_tracks, tg_playlist = ev.get_fold(dataset)
cbf.fit(urm, list(tg_playlist), list(tg_tracks), dataset)
recs = cbf.predict()
ev.evaluate_fold(recs)
map_at_five = ev.get_mean_map()
print("MAP@5 ", map_at_five)
def evaluateMap():
from src.utils.evaluator import Evaluator
dataset = Dataset(load_tags=True, filter_tag=False, weight_tag=False)
dataset.set_track_attr_weights_2(1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0,
0.0)
dataset.set_playlist_attr_weights(1, 1, 1, 0, 0)
ev = Evaluator()
ev.cross_validation(5, dataset.train_final.copy())
cbf = ContentBasedFiltering(alfa=0.999)
for i in range(0, 5):
urm, tg_tracks, tg_playlist = ev.get_fold(dataset)
cbf.fit(urm, list(tg_playlist), list(tg_tracks), dataset)
recs = cbf.predict()
ev.evaluate_fold(recs)
map_at_five = ev.get_mean_map()
print("MAP@5 ", map_at_five)
def run(self):
used_seconds = 0
# 迭代
while True:
logger.info(f"{'*' * 50}")
# 确定当前时间, 取算法执行时间和模拟器的切片时间的大值
self.cur_time = self.pre_time + (
used_seconds // self.time_interval + 1) * self.time_interval
logger.info(
f"cur time: {datetime.datetime.fromtimestamp(self.cur_time)}, "
f"pre time: {datetime.datetime.fromtimestamp(self.pre_time)}")
# update the status of vehicles and orders in a given interval [self.pre_time, self.cur_time]
updated_input_info = self.update_input()
# 派单环节, 设计与算法交互
used_seconds, dispatch_result = self.dispatch(updated_input_info)
self.time_to_dispatch_result[self.cur_time] = dispatch_result
# 校验, 车辆目的地不能改变
if not Checker.check_dispatch_result(
dispatch_result, self.id_to_vehicle, self.id_to_order):
logger.error("Dispatch result is infeasible")
return
# 根据派单指令更新车辆
self.deliver_control_command_to_vehicles(dispatch_result)
# 判断是否完成所有订单的派发
if self.complete_the_dispatch_of_all_orders():
break
self.pre_time = self.cur_time
# 若订单已经超时, 但是算法依旧未分配, 模拟终止
if self.ignore_allocating_timeout_orders(dispatch_result):
logger.error('Simulator terminated')
sys.exit(-1)
# 模拟完成车辆剩下的订单
self.simulate_the_left_ongoing_orders_of_vehicles(self.id_to_vehicle)
# 根据self.history 计算指标
self.total_score = Evaluator.calculate_total_score(
self.history, self.route_map, len(self.id_to_vehicle))
return self.R_hat
def get_model(self):
"""
Returns the complete R_hat
"""
return self.R_hat.copy()
if __name__ == '__main__':
ds = Dataset(load_tags=True, filter_tag=True)
ds.set_track_attr_weights_2(1,
1,
1,
1,
1,
num_rating_weight=1,
inferred_album=1,
inferred_duration=1,
inferred_playcount=1)
ev = Evaluator()
ev.cross_validation(5, ds.train_final.copy())
for i in range(0, 5):
fm = FM()
urm, tg_tracks, tg_playlist = ev.get_fold(ds)
fm.fit(urm, list(tg_playlist), list(tg_tracks), ds)
recs = fm.predict()
ev.evaluate_fold(recs)
map_at_five = ev.get_mean_map()
print("MAP@5 Final", map_at_five)