def getData():
p = 0.1
data = loadData(p=p)
data = featureEngineer(data)
# print(data.info())
#训练数据预处理
x_train, x_test, y_train, y_test = preprocessing(data)
return x_train, y_train, x_test, y_test
def predict_clf(model):
env = janestreet.make_env()
iter_test = env.iter_test()
for (test_df, sample_prediction_df) in iter_test:
if test_df['weight'].item() > 0:
test_df = featureEngineer(test_df)
X_test = test_df.loc[:, test_df.columns.str.contains('feature')]
X_test = X_test.fillna(0.0)
y_preds = model.predict(X_test)[0]
else:
y_preds = 0
# print(y_preds)
sample_prediction_df.action = y_preds
env.predict(sample_prediction_df)
print("%f (%f) with: %r" % (mean, stdev, param))
y.append(mean)
plt.plot(y)
plt.savefig("./output/" + param_name + ".png")
if __name__ == "__main__":
newpath = "/home/code"
os.chdir(newpath)
# data_explore()
# 真正开始干活
p = 0.0001
train = loadData(p=p)
train = featureEngineer(train)
# print(train.head())
# 计算模型评分
# score = Score(model, train)
# print("模型评分:%.2f" % score)
test = loadData(p=p)
test = featureEngineer(test)
#训练数据预处理
X_train, y_train = preprocessing(train)
# xgboost
print("XGBoost")
max_depth = [3, 4, 5]
subsample = [0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]
iter_test = env.iter_test()
for (test_df, sample_prediction_df) in iter_test:
if test_df['weight'].item() > 0:
X_test = test_df.loc[:, test_df.columns.str.contains('feature')]
X_test = X_test.fillna(-999)
y_preds = model.predict(X_test)[0]
else:
y_preds = 0
# print(y_preds)
sample_prediction_df.action = y_preds
env.predict(sample_prediction_df)
if __name__ == "__main__":
newpath = "/home/code"
os.chdir(newpath)
# data_explore()
# 真正开始干活
train = pd.read_csv("./train.csv", nrows=10000)
feature = pd.read_csv("./features.csv")
train = featureEngineer(train)
model = modeling(train)
# 计算模型评分
# score = Score(model, train)
# print("模型评分:%.2f" % score)
# 进行预测
predict_clf(model)