kaggle 入门 rossmann xgboost

对商店的某天进行销量预测

这个特征处理得很好，用上一个特征进行xgboost得到0.32的成绩，换成该特征得到0.14的成绩

特征：

Open为空赋为1

train中只看open为1且sales>0的记录

合并store和train&test

所有空值填0

'StoreType', 'Assortment', 'StateHoliday'中将0abcd变为01234

从日期中抽出Year, Month, Day, DOW, WOY

创建竞争对手开店月数，优惠月数，每条记录的所处月份是否是优惠月

几个特征的重要性为：Day>Store>竞争对手开店月数>WOY>DOW>优惠月数>CompetitionDistance>Month

代码：

#!/usr/bin/pythonimport csvimport operatorimport pandas as pdimport numpy as npfrom sklearn.cross_validation import train_test_splitimport xgboost as xgbimport matplotlibmatplotlib.use("Agg")  # Needed to save figuresimport matplotlib.pyplot as pltdef create_feature_map(features):outfile = open('xgb.fmap', 'w')for i, feat in enumerate(features):outfile.write('{0}\t{1}\tq\n'.format(i, feat))outfile.close()def rmspe(y, yhat):return np.sqrt(np.mean((yhat / y - 1) ** 2))def rmspe_xg(yhat, y):y = np.expm1(y.get_label())yhat = np.expm1(yhat)return "rmspe", rmspe(y, yhat)# Gather some featuresdef build_features(features, data):# remove NaNsdata.fillna(0, inplace=True)data.loc[data.Open.isnull(), 'Open'] = 1# Use some properties directlyfeatures.extend(['Store', 'CompetitionDistance', 'Promo', 'Promo2', 'SchoolHoliday'])# Label encode some featuresfeatures.extend(['StoreType', 'Assortment', 'StateHoliday'])mappings = {'0': 0, 'a': 1, 'b': 2, 'c': 3, 'd': 4}data.StoreType.replace(mappings, inplace=True)data.Assortment.replace(mappings, inplace=True)data.StateHoliday.replace(mappings, inplace=True)features.extend(['DayOfWeek', 'Month', 'Day', 'Year', 'WeekOfYear'])data['Year'] = data.Date.dt.yeardata['Month'] = data.Date.dt.monthdata['Day'] = data.Date.dt.daydata['DayOfWeek'] = data.Date.dt.dayofweekdata['WeekOfYear'] = data.Date.dt.weekofyear# CompetionOpen en PromoOpen from https://www.kaggle.com/ananya77041/rossmann-store-sales/randomforestpython/code# Calculate time competition open time in monthsfeatures.append('CompetitionOpen')data['CompetitionOpen'] = 12 * (data.Year - data.CompetitionOpenSinceYear) + \  (data.Month - data.CompetitionOpenSinceMonth)# Promo open time in monthsfeatures.append('PromoOpen')data['PromoOpen'] = 12 * (data.Year - data.Promo2SinceYear) + \(data.WeekOfYear - data.Promo2SinceWeek) / 4.0data['PromoOpen'] = data.PromoOpen.apply(lambda x: x if x > 0 else 0)data.loc[data.Promo2SinceYear == 0, 'PromoOpen'] = 0# Indicate that sales on that day are in promo intervalfeatures.append('IsPromoMonth')month2str = {1: 'Jan', 2: 'Feb', 3: 'Mar', 4: 'Apr', 5: 'May', 6: 'Jun', \ 7: 'Jul', 8: 'Aug', 9: 'Sept', 10: 'Oct', 11: 'Nov', 12: 'Dec'}data['monthStr'] = data.Month.map(month2str)data.loc[data.PromoInterval == 0, 'PromoInterval'] = ''data['IsPromoMonth'] = 0for interval in data.PromoInterval.unique():if interval != '':for month in interval.split(','):data.loc[(data.monthStr == month) & (data.PromoInterval == interval), 'IsPromoMonth'] = 1return data## Start of main scriptprint("Load the training, test and store data using pandas")types = {'CompetitionOpenSinceYear': np.dtype(int), 'CompetitionOpenSinceMonth': np.dtype(int), 'StateHoliday': np.dtype(str), 'Promo2SinceWeek': np.dtype(int), 'SchoolHoliday': np.dtype(float), 'PromoInterval': np.dtype(str)}train = pd.read_csv("../input/train.csv", parse_dates=[2], dtype=types)test = pd.read_csv("../input/test.csv", parse_dates=[3], dtype=types)store = pd.read_csv("../input/store.csv")print("Assume store open, if not provided")train.fillna(1, inplace=True)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("Use only Sales bigger then zero. Simplifies calculation of rmspe")train = train[train["Sales"] > 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)print('training data processed')params = {"objective": "reg:linear",  "booster": "gbtree",  "eta": 0.3,  "max_depth": 10,  "subsample": 0.9,  "colsample_bytree": 0.7,  "silent": 1,  "seed": 1301  }num_boost_round = 300print("Train a XGBoost model")X_train, X_valid = train_test_split(train, test_size=0.012, random_state=10)y_train = np.log1p(X_train.Sales)y_valid = np.log1p(X_valid.Sales)dtrain = xgb.DMatrix(X_train[features], y_train)dvalid = xgb.DMatrix(X_valid[features], y_valid)watchlist = [(dtrain, 'train'), (dvalid, 'eval')]gbm = xgb.train(params, dtrain, num_boost_round, evals=watchlist, \early_stopping_rounds=100, feval=rmspe_xg, verbose_eval=True)print("Validating")yhat = gbm.predict(xgb.DMatrix(X_valid[features]))error = rmspe(X_valid.Sales.values, np.expm1(yhat))print('RMSPE: {:.6f}'.format(error))print("Make predictions on the test set")dtest = xgb.DMatrix(test[features])test_probs = gbm.predict(dtest)# Make Submissionresult = pd.DataFrame({"Id": test["Id"], 'Sales': np.expm1(test_probs)})result.to_csv("xgboost_10_submission.csv", index=False)# XGB feature importances# Based on https://www.kaggle.com/mmueller/liberty-mutual-group-property-inspection-prediction/xgb-feature-importance-python/codecreate_feature_map(features)importance = gbm.get_fscore(fmap='xgb.fmap')importance = sorted(importance.items(), key=operator.itemgetter(1))df = pd.DataFrame(importance, columns=['feature', 'fscore'])df['fscore'] = df['fscore'] / df['fscore'].sum()featp = df.plot(kind='barh', x='feature', y='fscore', legend=False, figsize=(6, 10))plt.title('XGBoost Feature Importance')plt.xlabel('relative importance')fig_featp = featp.get_figure()fig_featp.savefig('feature_importance_xgb.png', bbox_inches='tight', pad_inches=1)

结果0.14