天池-零基础入门推荐系统-新闻推荐-排序模型和模型融合

排序

通过召回的操作， 我们已经进行了问题规模的缩减， 对于每个用户， 选择出了N篇文章作为了候选集，并基于召回的候选集构建了与用户历史相关的特征，以及用户本身的属性特征，文章本省的属性特征，以及用户与文章之间的特征，下面就是使用机器学习模型来对构造好的特征进行学习，然后对测试集进行预测，得到测试集中的每个候选集用户点击的概率，返回点击概率最大的topk个文章，作为最终的结果。

排序阶段选择了三个比较有代表性的排序模型，它们分别是：

LGB的排序模型
LGB的分类模型
深度学习的分类模型DIN

得到了最终的排序模型输出的结果之后，还选择了两种比较经典的模型集成的方法：

输出结果加权融合
Staking（将模型的输出结果再使用一个简单模型进行预测）

import numpy as np
import pandas as pd
import pickle
from tqdm import tqdm
import gc, os
import time
from datetime import datetime
import lightgbm as lgb
from sklearn.preprocessing import MinMaxScaler
import warnings
warnings.filterwarnings('ignore')
data_path = './data_raw/'
save_path = './temp_results/'
offline = False
# 重新读取数据的时候，发现click_article_id是一个浮点数，所以将其转换成int类型
trn_user_item_feats_df = pd.read_csv(save_path + 'trn_user_item_feats_df.csv')
trn_user_item_feats_df['click_article_id'] = trn_user_item_feats_df['click_article_id'].astype(int)

if offline:
    val_user_item_feats_df = pd.read_csv(save_path + 'val_user_item_feats_df.csv')
    val_user_item_feats_df['click_article_id'] = val_user_item_feats_df['click_article_id'].astype(int)
else:
    val_user_item_feats_df = None
    
tst_user_item_feats_df = pd.read_csv(save_path + 'tst_user_item_feats_df.csv')
tst_user_item_feats_df['click_article_id'] = tst_user_item_feats_df['click_article_id'].astype(int)

# 做特征的时候为了方便，给测试集也打上了一个无效的标签，这里直接删掉就行
del tst_user_item_feats_df['label']
def submit(recall_df, topk=5, model_name=None):
    recall_df = recall_df.sort_values(by=['user_id', 'pred_score'])
    recall_df['rank'] = recall_df.groupby(['user_id'])['pred_score'].rank(ascending=False, method='first')
    
    # 判断是不是每个用户都有5篇文章及以上
    tmp = recall_df.groupby('user_id').apply(lambda x: x['rank'].max())
    assert tmp.min() >= topk
    
    del recall_df['pred_score']
    submit = recall_df[recall_df['rank'] <= topk].set_index(['user_id', 'rank']).unstack(-1).reset_index()
    
    submit.columns = [int(col) if isinstance(col, int) else col for col in submit.columns.droplevel(0)]
    # 按照提交格式定义列名
    submit = submit.rename(columns={'': 'user_id', 1: 'article_1', 2: 'article_2', 
                                                  3: 'article_3', 4: 'article_4', 5: 'article_5'})
    
    save_name = save_path + model_name + '_' + datetime.today().strftime('%m-%d') + '.csv'
    submit.to_csv(save_name, index=False, header=True)
    def norm_sim(sim_df, weight=0.0):
    # print(sim_df.head())
    min_sim = sim_df.min()
    max_sim = sim_df.max()
    if max_sim == min_sim:
        sim_df = sim_df.apply(lambda sim: 1.0)
    else:
        sim_df = sim_df.apply(lambda sim: 1.0 * (sim - min_sim) / (max_sim - min_sim))

    sim_df = sim_df.apply(lambda sim: sim + weight)  # plus one
    return sim_df
    # 防止中间出错之后重新读取数据
trn_user_item_feats_df_rank_model = trn_user_item_feats_df.copy()

if offline:
    val_user_item_feats_df_rank_model = val_user_item_feats_df.copy()
    
tst_user_item_feats_df_rank_model = tst_user_item_feats_df.copy()
lgb_cols = ['sim0', 'time_diff0', 'word_diff0','sim_max', 'sim_min', 'sim_sum', 
            'sim_mean', 'score','click_size', 'time_diff_mean', 'active_level',
            'click_environment','click_deviceGroup', 'click_os', 'click_country', 
            'click_region','click_referrer_type', 'user_time_hob1', 'user_time_hob2',
            'words_hbo', 'category_id', 'created_at_ts','words_count']
            # 排序模型分组
trn_user_item_feats_df_rank_model.sort_values(by=['user_id'], inplace=True)
g_train = trn_user_item_feats_df_rank_model.groupby(['user_id'], as_index=False).count()["label"].values

if offline:
    val_user_item_feats_df_rank_model.sort_values(by=['user_id'], inplace=True)
    g_val = val_user_item_feats_df_rank_model.groupby(['user_id'], as_index=False).count()["label"].values
    # 排序模型定义
lgb_ranker = lgb.LGBMRanker(boosting_type='gbdt', num_leaves=31, reg_alpha=0.0, reg_lambda=1,
                            max_depth=-1, n_estimators=100, subsample=0.7, colsample_bytree=0.7, subsample_freq=1,
                            learning_rate=0.01, min_child_weight=50, random_state=2018, n_jobs= 16)  
                            # 排序模型训练
if offline:
    lgb_ranker.fit(trn_user_item_feats_df_rank_model[lgb_cols], trn_user_item_feats_df_rank_model['label'], group=g_train,
                eval_set=[(val_user_item_feats_df_rank_model[lgb_cols], val_user_item_feats_df_rank_model['label'])], 
                eval_group= [g_val], eval_at=[1, 2, 3, 4, 5], eval_metric=['ndcg', ], early_stopping_rounds=50, )
else:
    lgb_ranker.fit(trn_user_item_feats_df[lgb_cols], trn_user_item_feats_df['label'], group=g_train)
# 模型预测
tst_user_item_feats_df['pred_score'] = lgb_ranker.predict(tst_user_item_feats_df[lgb_cols], num_iteration=lgb_ranker.best_iteration_)

# 将这里的排序结果保存一份，用户后面的模型融合
tst_user_item_feats_df[['user_id', 'click_article_id', 'pred_score']].to_csv(save_path + 'lgb_ranker_score.csv', index=False)
# 预测结果重新排序, 及生成提交结果
rank_results = tst_user_item_feats_df[['user_id', 'click_article_id', 'pred_score']]
rank_results['click_article_id'] = rank_results['click_article_id'].astype(int)
submit(rank_results, topk=5, model_name='lgb_ranker')

# 五折交叉验证，这里的五折交叉是以用户为目标进行五折划分
#  这一部分与前面的单独训练和验证是分开的
def get_kfold_users(trn_df, n=5):
    user_ids = trn_df['user_id'].unique()
    user_set = [user_ids[i::n] for i in range(n)]
    return user_set

k_fold = 5
trn_df = trn_user_item_feats_df_rank_model
user_set = get_kfold_users(trn_df, n=k_fold)

score_list = []
score_df = trn_df[['user_id', 'click_article_id','label']]
sub_preds = np.zeros(tst_user_item_feats_df_rank_model.shape[0])

# 五折交叉验证，并将中间结果保存用于staking
for n_fold, valid_user in enumerate(user_set):
    train_idx = trn_df[~trn_df['user_id'].isin(valid_user)] # add slide user
    valid_idx = trn_df[trn_df['user_id'].isin(valid_user)]
    
    # 训练集与验证集的用户分组
    train_idx.sort_values(by=['user_id'], inplace=True)
    g_train = train_idx.groupby(['user_id'], as_index=False).count()["label"].values
    
    valid_idx.sort_values(by=['user_id'], inplace=True)
    g_val = valid_idx.groupby(['user_id'], as_index=False).count()["label"].values
    
    # 定义模型
    lgb_ranker = lgb.LGBMRanker(boosting_type='gbdt', num_leaves=31, reg_alpha=0.0, reg_lambda=1,
                            max_depth=-1, n_estimators=100, subsample=0.7, colsample_bytree=0.7, subsample_freq=1,
                            learning_rate=0.01, min_child_weight=50, random_state=2018, n_jobs= 16)  
    # 训练模型
    lgb_ranker.fit(train_idx[lgb_cols], train_idx['label'], group=g_train,
                   eval_set=[(valid_idx[lgb_cols], valid_idx['label'])], eval_group= [g_val], 
                   eval_at=[1, 2, 3, 4, 5], eval_metric=['ndcg', ], early_stopping_rounds=50, )
    
    # 预测验证集结果
    valid_idx['pred_score'] = lgb_ranker.predict(valid_idx[lgb_cols], num_iteration=lgb_ranker.best_iteration_)
    
    # 对输出结果进行归一化
    valid_idx['pred_score'] = valid_idx[['pred_score']].transform(lambda x: norm_sim(x))
    
    valid_idx.sort_values(by=['user_id', 'pred_score'])
    valid_idx['pred_rank'] = valid_idx.groupby(['user_id'])['pred_score'].rank(ascending=False, method='first')
    
    # 将验证集的预测结果放到一个列表中，后面进行拼接
    score_list.append(valid_idx[['user_id', 'click_article_id', 'pred_score', 'pred_rank']])
    
    # 如果是线上测试，需要计算每次交叉验证的结果相加，最后求平均
    if not offline:
        sub_preds += lgb_ranker.predict(tst_user_item_feats_df_rank_model[lgb_cols], lgb_ranker.best_iteration_)
    
score_df_ = pd.concat(score_list, axis=0)
score_df = score_df.merge(score_df_, how='left', on=['user_id', 'click_article_id'])
# 保存训练集交叉验证产生的新特征
score_df[['user_id', 'click_article_id', 'pred_score', 'pred_rank', 'label']].to_csv(save_path + 'trn_lgb_ranker_feats.csv', index=False)
    
# 测试集的预测结果，多次交叉验证求平均,将预测的score和对应的rank特征保存，可以用于后面的staking，这里还可以构造其他更多的特征
tst_user_item_feats_df_rank_model['pred_score'] = sub_preds / k_fold
tst_user_item_feats_df_rank_model['pred_score'] = tst_user_item_feats_df_rank_model['pred_score'].transform(lambda x: norm_sim(x))
tst_user_item_feats_df_rank_model.sort_values(by=['user_id', 'pred_score'])
tst_user_item_feats_df_rank_model['pred_rank'] = tst_user_item_feats_df_rank_model.groupby(['user_id'])['pred_score'].rank(ascending=False, method='first')

# 保存测试集交叉验证的新特征
tst_user_item_feats_df_rank_model[['user_id', 'click_article_id', 'pred_score', 'pred_rank']].to_csv(save_path + 'tst_lgb_ranker_feats.csv', index=False)

# 预测结果重新排序, 及生成提交结果
# 单模型生成提交结果
rank_results = tst_user_item_feats_df_rank_model[['user_id', 'click_article_id', 'pred_score']]
rank_results['click_article_id'] = rank_results['click_article_id'].astype(int)
submit(rank_results, topk=5, model_name='lgb_ranker')

# 模型及参数的定义
lgb_Classfication = lgb.LGBMClassifier(boosting_type='gbdt', num_leaves=31, reg_alpha=0.0, reg_lambda=1,
                            max_depth=-1, n_estimators=500, subsample=0.7, colsample_bytree=0.7, subsample_freq=1,
                            learning_rate=0.01, min_child_weight=50, random_state=2018, n_jobs= 16, verbose=10)  
# 模型训练
if offline:
    lgb_Classfication.fit(trn_user_item_feats_df_rank_model[lgb_cols], trn_user_item_feats_df_rank_model['label'],
                    eval_set=[(val_user_item_feats_df_rank_model[lgb_cols], val_user_item_feats_df_rank_model['label'])], 
                    eval_metric=['auc', ],early_stopping_rounds=50, )
else:
    lgb_Classfication.fit(trn_user_item_feats_df_rank_model[lgb_cols], trn_user_item_feats_df_rank_model['label'])
# 五折交叉验证，这里的五折交叉是以用户为目标进行五折划分
#  这一部分与前面的单独训练和验证是分开的
def get_kfold_users(trn_df, n=5):
    user_ids = trn_df['user_id'].unique()
    user_set = [user_ids[i::n] for i in range(n)]
    return user_set

k_fold = 5
trn_df = trn_user_item_feats_df_rank_model
user_set = get_kfold_users(trn_df, n=k_fold)

score_list = []
score_df = trn_df[['user_id', 'click_article_id', 'label']]
sub_preds = np.zeros(tst_user_item_feats_df_rank_model.shape[0])

# 五折交叉验证，并将中间结果保存用于staking
for n_fold, valid_user in enumerate(user_set):
    train_idx = trn_df[~trn_df['user_id'].isin(valid_user)] # add slide user
    valid_idx = trn_df[trn_df['user_id'].isin(valid_user)]
    
    # 模型及参数的定义
    lgb_Classfication = lgb.LGBMClassifier(boosting_type='gbdt', num_leaves=31, reg_alpha=0.0, reg_lambda=1,
                            max_depth=-1, n_estimators=100, subsample=0.7, colsample_bytree=0.7, subsample_freq=1,
                            learning_rate=0.01, min_child_weight=50, random_state=2018, n_jobs= 16, verbose=10)  
    # 训练模型
    lgb_Classfication.fit(train_idx[lgb_cols], train_idx['label'],eval_set=[(valid_idx[lgb_cols], valid_idx['label'])], 
                          eval_metric=['auc', ],early_stopping_rounds=50, )
    
    # 预测验证集结果
    valid_idx['pred_score'] = lgb_Classfication.predict_proba(valid_idx[lgb_cols], 
                                                              num_iteration=lgb_Classfication.best_iteration_)[:,1]
    
    # 对输出结果进行归一化 分类模型输出的值本身就是一个概率值不需要进行归一化
    # valid_idx['pred_score'] = valid_idx[['pred_score']].transform(lambda x: norm_sim(x))
    
    valid_idx.sort_values(by=['user_id', 'pred_score'])
    valid_idx['pred_rank'] = valid_idx.groupby(['user_id'])['pred_score'].rank(ascending=False, method='first')
    
    # 将验证集的预测结果放到一个列表中，后面进行拼接
    score_list.append(valid_idx[['user_id', 'click_article_id', 'pred_score', 'pred_rank']])
    
    # 如果是线上测试，需要计算每次交叉验证的结果相加，最后求平均
    if not offline:
        sub_preds += lgb_Classfication.predict_proba(tst_user_item_feats_df_rank_model[lgb_cols], 
                                                     num_iteration=lgb_Classfication.best_iteration_)[:,1]
    
score_df_ = pd.concat(score_list, axis=0)
score_df = score_df.merge(score_df_, how='left', on=['user_id', 'click_article_id'])
# 保存训练集交叉验证产生的新特征
score_df[['user_id', 'click_article_id', 'pred_score', 'pred_rank', 'label']].to_csv(save_path + 'trn_lgb_cls_feats.csv', index=False)
    
# 测试集的预测结果，多次交叉验证求平均,将预测的score和对应的rank特征保存，可以用于后面的staking，这里还可以构造其他更多的特征
tst_user_item_feats_df_rank_model['pred_score'] = sub_preds / k_fold
tst_user_item_feats_df_rank_model['pred_score'] = tst_user_item_feats_df_rank_model['pred_score'].transform(lambda x: norm_sim(x))
tst_user_item_feats_df_rank_model.sort_values(by=['user_id', 'pred_score'])
tst_user_item_feats_df_rank_model['pred_rank'] = tst_user_item_feats_df_rank_model.groupby(['user_id'])['pred_score'].rank(ascending=False, method='first')

# 保存测试集交叉验证的新特征
tst_user_item_feats_df_rank_model[['user_id', 'click_article_id', 'pred_score', 'pred_rank']].to_csv(save_path + 'tst_lgb_cls_feats.csv', index=False)

# 预测结果重新排序, 及生成提交结果
rank_results = tst_user_item_feats_df_rank_model[['user_id', 'click_article_id', 'pred_score']]
rank_results['click_article_id'] = rank_results['click_article_id'].astype(int)
submit(rank_results, topk=5, model_name='lgb_cls')

模型融合

加权融合

# 读取多个模型的排序结果文件
lgb_ranker = pd.read_csv(save_path + 'lgb_ranker_score.csv')
lgb_cls = pd.read_csv(save_path + 'lgb_cls_score.csv')
din_ranker = pd.read_csv(save_path + 'din_rank_score.csv')

# 这里也可以换成交叉验证输出的测试结果进行加权融合
rank_model = {'lgb_ranker': lgb_ranker, 
              'lgb_cls': lgb_cls, 
              'din_ranker': din_ranker}
def get_ensumble_predict_topk(rank_model, topk=5):
    final_recall = rank_model['lgb_cls'].append(rank_model['din_ranker'])
    rank_model['lgb_ranker']['pred_score'] = rank_model['lgb_ranker']['pred_score'].transform(lambda x: norm_sim(x))
    
    final_recall = final_recall.append(rank_model['lgb_ranker'])
    final_recall = final_recall.groupby(['user_id', 'click_article_id'])['pred_score'].sum().reset_index()
    
    submit(final_recall, topk=topk, model_name='ensemble_fuse')
get_ensumble_predict_topk(rank_model)

staking

# 读取多个模型的交叉验证生成的结果文件
# 训练集
trn_lgb_ranker_feats = pd.read_csv(save_path + 'trn_lgb_ranker_feats.csv')
trn_lgb_cls_feats = pd.read_csv(save_path + 'trn_lgb_cls_feats.csv')
trn_din_cls_feats = pd.read_csv(save_path + 'trn_din_cls_feats.csv')

# 测试集
tst_lgb_ranker_feats = pd.read_csv(save_path + 'tst_lgb_ranker_feats.csv')
tst_lgb_cls_feats = pd.read_csv(save_path + 'tst_lgb_cls_feats.csv')
tst_din_cls_feats = pd.read_csv(save_path + 'tst_din_cls_feats.csv')

# 将多个模型输出的特征进行拼接

finall_trn_ranker_feats = trn_lgb_ranker_feats[['user_id', 'click_article_id', 'label']]
finall_tst_ranker_feats = tst_lgb_ranker_feats[['user_id', 'click_article_id']]

for idx, trn_model in enumerate([trn_lgb_ranker_feats, trn_lgb_cls_feats, trn_din_cls_feats]):
    for feat in [ 'pred_score', 'pred_rank']:
        col_name = feat + '_' + str(idx)
        finall_trn_ranker_feats[col_name] = trn_model[feat]

for idx, tst_model in enumerate([tst_lgb_ranker_feats, tst_lgb_cls_feats, tst_din_cls_feats]):
    for feat in [ 'pred_score', 'pred_rank']:
        col_name = feat + '_' + str(idx)
        finall_tst_ranker_feats[col_name] = tst_model[feat]
# 定义一个逻辑回归模型再次拟合交叉验证产生的特征对测试集进行预测
# 这里需要注意的是，在做交叉验证的时候可以构造多一些与输出预测值相关的特征，来丰富这里简单模型的特征
from sklearn.linear_model import LogisticRegression

feat_cols = ['pred_score_0', 'pred_rank_0', 'pred_score_1', 'pred_rank_1', 'pred_score_2', 'pred_rank_2']

trn_x = finall_trn_ranker_feats[feat_cols]
trn_y = finall_trn_ranker_feats['label']

tst_x = finall_tst_ranker_feats[feat_cols]

# 定义模型
lr = LogisticRegression()

# 模型训练
lr.fit(trn_x, trn_y)

# 模型预测
finall_tst_ranker_feats['pred_score'] = lr.predict_proba(tst_x)[:, 1]

# 预测结果重新排序, 及生成提交结果
rank_results = finall_tst_ranker_feats[['user_id', 'click_article_id', 'pred_score']]
submit(rank_results, topk=5, model_name='ensumble_staking')

总结

本章主要学习了三个排序模型，包括LGB的Rank， LGB的Classifier还有深度学习的DIN模型， 当然，对于这三个模型的原理部分，我们并没有给出详细的介绍， 请大家课下自己探索原理，也欢迎大家把自己的探索与所学分享出来，我们一块学习和进步。最后，我们进行了简单的模型融合策略，包括简单的加权和Stacking。