class NnStackedDt:
nn_model = []
dt_model = []
def __init__(self):
self.nn_model = DeepNeuralNetwork(INPUT_DIMS_MQ2008)
self.dt_model = DecisionTree(100)
def fit(self, train_data):
self.nn_model.fit(train_data)
pred = self.nn_model.predict(train_data)
stacked_data = train_data
stacked_data[0] = pred
self.dt_model.fit(stacked_data)
def predict(self, test_data):
pred = self.nn_model.predict(test_data)
stacked_data = test_data
stacked_data[0] = pred
return self.dt_model.predict(stacked_data)
class NnBoostedDt:
nn_model = []
dt_model = []
def __init__(self):
self.nn_model = DeepNeuralNetwork(INPUT_DIMS_MQ2008)
self.dt_model = AdaBoostClassifier(n_estimators=300, random_state=0)
def fit(self, train_data):
self.nn_model.fit(train_data)
pred = self.nn_model.predict(train_data)
train_x, train_y, train_q = get_data_params(train_data)
boosted_data = train_data
boosted_data[0] = pred
self.dt_model.fit(boosted_data, np.subtract(pred, train_y))
def predict(self, test_data):
pred_nn = self.nn_model.predict(test_data)
pred_dt = self.dt_model.predict(test_data)
return pred_nn + pred_dt
class DtStackedNn:
nn_model = []
dt_model = []
def __init__(self):
self.nn_model = DeepNeuralNetwork(INPUT_DIMS_MQ2008 + 1)
self.dt_model = DecisionTree(200)
def fit(self, train_data):
self.dt_model.fit(train_data)
pred = self.dt_model.predict(train_data)
stacked_data = train_data
stacked_data[len(stacked_data.columns)] = pred
self.nn_model.fit(stacked_data)
def predict(self, test_data):
pred = self.dt_model.predict(test_data)
stacked_data = test_data
stacked_data[len(stacked_data.columns)] = pred
return self.nn_model.predict(stacked_data)
class NnDtBagger:
nn_model = []
dt_model = []
nn_weight = 0.5
def __init__(self, nn_weight):
self.nn_weight = nn_weight
self.nn_model = DeepNeuralNetwork(INPUT_DIMS_MQ2008)
self.dt_model = DecisionTree(100)
def fit(self, train_data):
# TODO: do some data processing if needed (data division)
self.nn_model.fit(train_data)
self.dt_model.fit(train_data)
def predict(self, test_data):
pred = self.nn_weight * np.asarray(self.nn_model.predict(
test_data)) + (1 - self.nn_weight) * np.asarray(
self.dt_model.predict(test_data))
return pred.astype(int)
def __init__(self):
self.nn_model = DeepNeuralNetwork(INPUT_DIMS_MQ2008)
self.dt_model = AdaBoostClassifier(n_estimators=300, random_state=0)
def __init__(self):
self.nn_model = DeepNeuralNetwork(INPUT_DIMS_MQ2008)
self.dt_model = DecisionTree(100)
from sklearn.model_selection import train_test_split
from util.Utils import read_dataset_as_df
from constant.Constant import DATASET_MQ2008_PATH
from constant.Constant import MQ2008_TSV_FILE_NAME
from model.NeuralNetwork import DeepNeuralNetwork
from util.Utils import get_data_params
from util.Utils import calculate_ndcg
from util.Utils import calculate_map
from constant.Constant import INPUT_DIMS_MQ2008
''''
Main Running logic for the Deep NN model with out ensemble
Dataset : MQ2008
'''
print(":: Main Running logic for the Deep NN model with out ensemble ::")
df = read_dataset_as_df(DATASET_MQ2008_PATH + MQ2008_TSV_FILE_NAME)
# Dividing the data
train, test = train_test_split(df, test_size=0.7)
train_x, train_y, train_q = get_data_params(train)
test_x, test_y, test_q = get_data_params(test)
model = DeepNeuralNetwork(INPUT_DIMS_MQ2008)
model.fit(train)
pred = model.predict(test)
ndcg = calculate_ndcg(pred, test_y)
mAP = calculate_map(pred, test_y)
print('NDCG For Deep NN :', ndcg)
print('MAP For Deep NN :', mAP)