

In [1]:

    
#forked from Gilberto Titericz Junior


import pandas as pd
import numpy as np
from sklearn import ensemble, preprocessing
import xgboost as xgb

# load training and test datasets
train = pd.read_csv('../competition_data/train_set.csv', parse_dates=[2,])
test = pd.read_csv('../competition_data/test_set.csv', parse_dates=[3,])
tube_data = pd.read_csv('../competition_data/tube.csv')
bill_of_materials_data = pd.read_csv('../competition_data/bill_of_materials.csv')
specs_data = pd.read_csv('../competition_data/specs.csv')

# print("train columns")
# print(train.columns)
# print("test columns")
# print(test.columns)
# print("tube.csv df columns")
# print(tube_data.columns)
# print("bill_of_materials.csv df columns")
# print(bill_of_materials_data.columns)
# print("specs.csv df columns")
# print(specs_data.columns)


# print(specs_data[2:3])

train = pd.merge(train, tube_data, on ='tube_assembly_id')
train = pd.merge(train, bill_of_materials_data, on ='tube_assembly_id')
test = pd.merge(test, tube_data, on ='tube_assembly_id')
test = pd.merge(test, bill_of_materials_data, on ='tube_assembly_id')

# print("new train columns")
# print(train.columns)
# print(train[1:10])
# print(train.columns.to_series().groupby(train.dtypes).groups)

# create some new features
train['year'] = train.quote_date.dt.year
train['month'] = train.quote_date.dt.month
#train['dayofyear'] = train.quote_date.dt.dayofyear
#train['dayofweek'] = train.quote_date.dt.dayofweek
#train['day'] = train.quote_date.dt.day

test['year'] = test.quote_date.dt.year
test['month'] = test.quote_date.dt.month
#test['dayofyear'] = test.quote_date.dt.dayofyear
#test['dayofweek'] = test.quote_date.dt.dayofweek
#test['day'] = test.quote_date.dt.day

# drop useless columns and create labels
idx = test.id.values.astype(int)
test = test.drop(['id', 'tube_assembly_id', 'quote_date'], axis = 1)
labels = train.cost.values
#'tube_assembly_id', 'supplier', 'bracket_pricing', 'material_id', 'end_a_1x', 'end_a_2x', 'end_x_1x', 'end_x_2x', 'end_a', 'end_x'
#for some reason material_id cannot be converted to categorical variable
train = train.drop(['quote_date', 'cost', 'tube_assembly_id'], axis = 1)

train['material_id'].replace(np.nan,' ', regex=True, inplace= True)
test['material_id'].replace(np.nan,' ', regex=True, inplace= True)
for i in range(1,9):
    column_label = 'component_id_'+str(i)
    print(column_label)
    train[column_label].replace(np.nan,' ', regex=True, inplace= True)
    test[column_label].replace(np.nan,' ', regex=True, inplace= True)

train.fillna(0, inplace = True)
test.fillna(0, inplace = True)

print("train columns")
print(train.columns)

# convert data to numpy array
train = np.array(train)
test = np.array(test)

# label encode the categorical variables
for i in range(train.shape[1]):
    if i in [0,3,5,11,12,13,14,15,16,20,22,24,26,28,30,32,34]:
        print(i,list(train[1:5,i]) + list(test[1:5,i]))
        lbl = preprocessing.LabelEncoder()
        lbl.fit(list(train[:,i]) + list(test[:,i]))
        train[:,i] = lbl.transform(train[:,i])
        test[:,i] = lbl.transform(test[:,i])


# object array to float
train = train.astype(float)
test = test.astype(float)

# i like to train on log(1+x) for RMSLE ;) 
# The choice is yours :)
label_log = np.log1p(labels)

# fit a random forest model

params = {}
params["objective"] = "reg:linear"
params["eta"] = 0.02
params["min_child_weight"] = 5
params["subsample"] = 0.7
params["colsample_bytree"] = 0.6
params["scale_pos_weight"] = 0.8
params["silent"] = 1
params["max_depth"] = 9
params["max_delta_step"]=2

plst = list(params.items())

xgtrain = xgb.DMatrix(train, label=label_log)
xgtest = xgb.DMatrix(test)


num_rounds = 2000
model = xgb.train(plst, xgtrain, num_rounds)
preds1 = model.predict(xgtest)


num_rounds = 3000
model = xgb.train(plst, xgtrain, num_rounds)
preds2 = model.predict(xgtest)

num_rounds = 1500
model = xgb.train(plst, xgtrain, num_rounds)
preds4 = model.predict(xgtest)

label_log = np.power(labels,1.0/16)

xgtrain = xgb.DMatrix(train, label=label_log)
xgtest = xgb.DMatrix(test)
num_rounds = 4000
model = xgb.train(plst, xgtrain, num_rounds)
preds3 = model.predict(xgtest)

#for loop in range(2):
#    model = xgb.train(plst, xgtrain, num_rounds)
#    preds1 = preds1 + model.predict(xgtest)
preds = (np.expm1( (preds1+preds2+preds4)/3)+np.power(preds3,16))/2

preds = pd.DataFrame({"id": idx, "cost": preds})
# preds.to_csv('benchmark.csv', index=False)









    



component_id_1
component_id_2
component_id_3
component_id_4
component_id_5
component_id_6
component_id_7
component_id_8
train columns
Index([u'supplier', u'annual_usage', u'min_order_quantity', u'bracket_pricing',
       u'quantity', u'material_id', u'diameter', u'wall', u'length',
       u'num_bends', u'bend_radius', u'end_a_1x', u'end_a_2x', u'end_x_1x',
       u'end_x_2x', u'end_a', u'end_x', u'num_boss', u'num_bracket', u'other',
       u'component_id_1', u'quantity_1', u'component_id_2', u'quantity_2',
       u'component_id_3', u'quantity_3', u'component_id_4', u'quantity_4',
       u'component_id_5', u'quantity_5', u'component_id_6', u'quantity_6',
       u'component_id_7', u'quantity_7', u'component_id_8', u'quantity_8',
       u'year', u'month'],
      dtype='object')
(0, ['S-0066', 'S-0066', 'S-0066', 'S-0066', 'S-0066', 'S-0066', 'S-0066', 'S-0066'])
(3, ['Yes', 'Yes', 'Yes', 'Yes', 'Yes', 'Yes', 'Yes', 'Yes'])
(5, ['SP-0019', 'SP-0019', 'SP-0019', 'SP-0019', 'SP-0035', 'SP-0035', 'SP-0035', 'SP-0035'])
(11, ['N', 'N', 'N', 'N', 'N', 'N', 'N', 'N'])
(12, ['N', 'N', 'N', 'N', 'N', 'N', 'N', 'N'])
(13, ['N', 'N', 'N', 'N', 'N', 'N', 'N', 'N'])
(14, ['N', 'N', 'N', 'N', 'N', 'N', 'N', 'N'])
(15, ['EF-008', 'EF-008', 'EF-008', 'EF-008', 'EF-003', 'EF-003', 'EF-003', 'EF-003'])
(16, ['EF-008', 'EF-008', 'EF-008', 'EF-008', 'EF-003', 'EF-003', 'EF-003', 'EF-003'])
(20, ['C-1312', 'C-1312', 'C-1312', 'C-1312', 'C-1622', 'C-1622', 'C-1622', 'C-1622'])
(22, [' ', ' ', ' ', ' ', 'C-1629', 'C-1629', 'C-1629', 'C-1629'])
(24, [' ', ' ', ' ', ' ', ' ', ' ', ' ', ' '])
(26, [' ', ' ', ' ', ' ', ' ', ' ', ' ', ' '])
(28, [' ', ' ', ' ', ' ', ' ', ' ', ' ', ' '])
(30, [' ', ' ', ' ', ' ', ' ', ' ', ' ', ' '])
(32, [' ', ' ', ' ', ' ', ' ', ' ', ' ', ' '])
(34, [' ', ' ', ' ', ' ', ' ', ' ', ' ', ' '])