learning.py

print("Load the training, test and store data using pandas")
train = pd.read_csv("input/train.csv")
test = pd.read_csv("input/test.csv")
store = pd.read_csv("input/store.csv")

print("Assume store open, if not provided")
test.fillna(1, inplace=True)

print("Consider only open stores for training. Closed stores wont count into the score.")
train = train[train["Open"] != 0]

print("Join with store")
train = pd.merge(train, store, on='Store')
test = pd.merge(test, store, on='Store')

features = []

print("augment features")
build_features(features, train)
build_features([], test)
print(features)

params = {"objective": "reg:linear",
          "eta": 0.3,
          "max_depth": 8,
          "subsample": 0.7,
          "colsample_bytree": 0.7,
          "silent": 1
          }
num_trees = 300

print("Train a XGBoost model")
val_size = 100000
#train = train.sort(['Date'])
print(train.tail(1)['Date'])
X_train, X_test = cross_validation.train_test_split(train, test_size=0.01)
#X_train, X_test = train.head(len(train) - val_size), train.tail(val_size)
dtrain = xgb.DMatrix(X_train[features], np.log(X_train["Sales"] + 1))
dvalid = xgb.DMatrix(X_test[features], np.log(X_test["Sales"] + 1))
dtest = xgb.DMatrix(test[features])
watchlist = [(dvalid, 'eval'), (dtrain, 'train')]
gbm = xgb.train(params, dtrain, num_trees, evals=watchlist, early_stopping_rounds=50, feval=rmspe_xg, verbose_eval=True)

gbm.save_model('model\\base.model')

print("Validating")
train_probs = gbm.predict(xgb.DMatrix(X_test[features]))
indices = train_probs < 0
train_probs[indices] = 0
error = rmspe(np.exp(train_probs) - 1, X_test['Sales'].values)
print('error', error)

print("Make predictions on the test set")
test_probs = gbm.predict(xgb.DMatrix(test[features]))
indices = test_probs < 0
test_probs[indices] = 0
submission = pd.DataFrame({"Id": test["Id"], "Sales": np.exp(test_probs) - 1})
submission.to_csv("output\\base_submission.csv", index=False)