讯飞电力需求预测挑战赛上分笔记 Datawhale AI 夏令营

1 开始

比赛地址：2024 iFLYTEK A.I.开发者大赛-讯飞开放平台

datawhale夏令营手册：Docs

百度飞浆平台：飞桨AI Studio星河社区-人工智能学习与实训社区

以及别忘了我写过的一篇：飞桨PaddlePaddle平台算力白嫖从入门到入狱不完全指南（Datawhale AI 夏令营）-CSDN博客

2 lightgbm 闭着眼跑：259分

2.1 环境配置与文件导入

!pip install lightgbm==3.3.0
import numpy as np
import pandas as pd
import lightgbm as lgb
from sklearn.metrics import mean_squared_log_error, mean_absolute_error, mean_squared_error
import tqdm
import sys
import os
import gc
import argparse
import warnings
warnings.filterwarnings('ignore')

train = pd.read_csv('./data/train.csv')
test = pd.read_csv('./data/test.csv')

2.2 特征工程（有修改）

# 合并训练数据和测试数据，并进行排序
data = pd.concat([test, train], axis=0, ignore_index=True)
data = data.sort_values(['id','dt'], ascending=False).reset_index(drop=True)

# 历史平移
for i in range(10,30):
    data[f'last{i}_target'] = data.groupby(['id'])['target'].shift(i)
    
# 窗口统计
data[f'win3_mean_target'] = (data['last10_target'] + data['last11_target'] + data['last12_target']) / 3

# 进行数据切分
train = data[data.target.notnull()].reset_index(drop=True)
test = data[data.target.isnull()].reset_index(drop=True)

# 确定输入特征
train_cols = [f for f in data.columns if f not in ['id','target']]

2.3 跑啊

def time_model(lgb, train_df, test_df, cols):
    # 训练集和验证集切分
    trn_x, trn_y = train_df[train_df.dt>=31][cols], train_df[train_df.dt>=31]['target']
    val_x, val_y = train_df[train_df.dt<=30][cols], train_df[train_df.dt<=30]['target']
    # 构建模型输入数据
    train_matrix = lgb.Dataset(trn_x, label=trn_y)
    valid_matrix = lgb.Dataset(val_x, label=val_y)
    # lightgbm参数
    lgb_params = {
        'boosting_type': 'gbdt',
        'objective': 'regression',
        'metric': 'mse',
        'min_child_weight': 5,
        'num_leaves': 2 ** 5,
        'lambda_l2': 10,
        'feature_fraction': 0.8,
        'bagging_fraction': 0.8,
        'bagging_freq': 4,
        'learning_rate': 0.05,
        'seed': 2024,
        'nthread' : 16,
        'verbose' : -1,
    }

    # 训练模型
    model = lgb.train(lgb_params, train_matrix, 50000, valid_sets=[train_matrix, valid_matrix], 
                      categorical_feature=[], verbose_eval=500, early_stopping_rounds=500)
    # from lightgbm.callback import log_evaluation
    # model = lgb.train(lgb_params, train_matrix, 50000, valid_sets=[train_matrix, valid_matrix],
    #               callbacks=[log_evaluation(period=100)])
    # 验证集和测试集结果预测
    val_pred = model.predict(val_x, num_iteration=model.best_iteration)
    test_pred = model.predict(test_df[cols], num_iteration=model.best_iteration)
    # 离线分数评估
    score = mean_squared_error(val_pred, val_y)
    print(score)
       
    return val_pred, test_pred
    
lgb_oof, lgb_test = time_model(lgb, train, test, train_cols)

# 保存结果文件到本地
test['target'] = lgb_test
test[['id','dt','target']].to_csv('submit2.csv', index=None)

3 模型融合：236分

3.1 环境配置

!pip install lightgbm==3.3.0
!pip install xgboost -i https://pypi.tuna.tsinghua.edu.cn/simple # 指定清华镜像
!pip install catboost -i https://pypi.tuna.tsinghua.edu.cn/simple --user

3.2 特征工程

import numpy as np
import pandas as pd
import lightgbm as lgb
from sklearn.metrics import mean_squared_log_error, mean_absolute_error, mean_squared_error
import tqdm
import sys
import os
import gc
import argparse
import warnings
warnings.filterwarnings('ignore')

train = pd.read_csv('./data/train.csv')
test = pd.read_csv('./data/test.csv')

# 合并训练数据和测试数据，并进行排序
data = pd.concat([test, train], axis=0, ignore_index=True)
data = data.sort_values(['id','dt'], ascending=False).reset_index(drop=True)

# 历史平移
for i in range(10,36):
    data[f'target_shift{i}'] = data.groupby('id')['target'].shift(i)

# 历史平移 + 差分特征
for i in range(1,4):
    data[f'target_shift10_diff{i}'] = data.groupby('id')['target_shift10'].diff(i)
    
# 窗口统计
for win in [15,30,50,70]:
    data[f'target_win{win}_mean'] = data.groupby('id')['target'].rolling(window=win, min_periods=3, closed='left').mean().values
    data[f'target_win{win}_max'] = data.groupby('id')['target'].rolling(window=win, min_periods=3, closed='left').max().values
    data[f'target_win{win}_min'] = data.groupby('id')['target'].rolling(window=win, min_periods=3, closed='left').min().values
    data[f'target_win{win}_std'] = data.groupby('id')['target'].rolling(window=win, min_periods=3, closed='left').std().values

# 历史平移 + 窗口统计
for win in [7,14,28,35,50,70]:
    data[f'target_shift10_win{win}_mean'] = data.groupby('id')['target_shift10'].rolling(window=win, min_periods=3, closed='left').mean().values
    data[f'target_shift10_win{win}_max'] = data.groupby('id')['target_shift10'].rolling(window=win, min_periods=3, closed='left').max().values
    data[f'target_shift10_win{win}_min'] = data.groupby('id')['target_shift10'].rolling(window=win, min_periods=3, closed='left').min().values
    data[f'target_shift10_win{win}_sum'] = data.groupby('id')['target_shift10'].rolling(window=win, min_periods=3, closed='left').sum().values
    data[f'target_shift710win{win}_std'] = data.groupby('id')['target_shift10'].rolling(window=win, min_periods=3, closed='left').std().values

# 进行数据切分
train = data[data.target.notnull()].reset_index(drop=True)
test = data[data.target.isnull()].reset_index(drop=True)

# 确定输入特征
train_cols = [f for f in data.columns if f not in ['id','target']]

3.3 解决爆内存

出现这个问题是内存炸了

解决方案1：更大的内存！！！

算力不够用可以参考我的《白嫖指南》

解决方案2：将浮点型数据float64转换为float32或float16

参考Datawhale AI 夏令营-电力需求预测挑战赛 · 语雀

回头我再试试

3.4 跑起来了！！！

from sklearn.model_selection import StratifiedKFold, KFold, GroupKFold
import lightgbm as lgb
import xgboost as xgb
from catboost import CatBoostRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
def cv_model(clf, train_x, train_y, test_x, clf_name, seed = 2024):
    '''
    clf：调用模型
    train_x：训练数据
    train_y：训练数据对应标签
    test_x：测试数据
    clf_name：选择使用模型名
    seed：随机种子
    '''
    folds = 5
    kf = KFold(n_splits=folds, shuffle=True, random_state=seed)
    oof = np.zeros(train_x.shape[0])
    test_predict = np.zeros(test_x.shape[0])
    cv_scores = []
    
    for i, (train_index, valid_index) in enumerate(kf.split(train_x, train_y)):
        print('************************************ {} ************************************'.format(str(i+1)))
        trn_x, trn_y, val_x, val_y = train_x.iloc[train_index], train_y[train_index], train_x.iloc[valid_index], train_y[valid_index]
        
        if clf_name == "lgb":
            train_matrix = clf.Dataset(trn_x, label=trn_y)
            valid_matrix = clf.Dataset(val_x, label=val_y)
            params = {
                'boosting_type': 'gbdt',
                'objective': 'regression',
                'metric': 'mae',
                'min_child_weight': 6,
                'num_leaves': 2 ** 6,
                'lambda_l2': 10,
                'feature_fraction': 0.8,
                'bagging_fraction': 0.8,
                'bagging_freq': 4,
                'learning_rate': 0.1,
                'seed': 2023,
                'nthread' : 16,
                'verbose' : -1,
            }
            model = clf.train(params, train_matrix, 1000, valid_sets=[train_matrix, valid_matrix],
                              categorical_feature=[], verbose_eval=200, early_stopping_rounds=100)
            val_pred = model.predict(val_x, num_iteration=model.best_iteration)
            test_pred = model.predict(test_x, num_iteration=model.best_iteration)
        
        if clf_name == "xgb":
            xgb_params = {
              'booster': 'gbtree', 
              'objective': 'reg:squarederror',
              'eval_metric': 'mae',
              'max_depth': 5,
              'lambda': 10,
              'subsample': 0.7,
              'colsample_bytree': 0.7,
              'colsample_bylevel': 0.7,
              'eta': 0.1,
              'tree_method': 'hist',
              'seed': 520,
              'nthread': 16
              }
            train_matrix = clf.DMatrix(trn_x , label=trn_y)
            valid_matrix = clf.DMatrix(val_x , label=val_y)
            test_matrix = clf.DMatrix(test_x)
            
            watchlist = [(train_matrix, 'train'),(valid_matrix, 'eval')]
            
            model = clf.train(xgb_params, train_matrix, num_boost_round=1000, evals=watchlist, verbose_eval=200, early_stopping_rounds=100)
            val_pred  = model.predict(valid_matrix)
            test_pred = model.predict(test_matrix)
            
        if clf_name == "cat":
            params = {'learning_rate': 0.1, 'depth': 5, 'bootstrap_type':'Bernoulli','random_seed':2023,
                      'od_type': 'Iter', 'od_wait': 100, 'random_seed': 11, 'allow_writing_files': False}
            
            model = clf(iterations=1000, **params)
            model.fit(trn_x, trn_y, eval_set=(val_x, val_y),
                      metric_period=200,
                      use_best_model=True, 
                      cat_features=[],
                      verbose=1)
            
            val_pred  = model.predict(val_x)
            test_pred = model.predict(test_x)
        
        oof[valid_index] = val_pred
        test_predict += test_pred / kf.n_splits
        
        score = mean_absolute_error(val_y, val_pred)
        cv_scores.append(score)
        print(cv_scores)
        
    return oof, test_predict

# 选择lightgbm模型
lgb_oof, lgb_test = cv_model(lgb, train[train_cols], train['target'], test[train_cols], 'lgb')
# 选择xgboost模型
xgb_oof, xgb_test = cv_model(xgb, train[train_cols], train['target'], test[train_cols], 'xgb')
# 选择catboost模型
cat_oof, cat_test = cv_model(CatBoostRegressor, train[train_cols], train['target'], test[train_cols], 'cat')

# 进行取平均融合
final_test = (lgb_test + xgb_test + cat_test) / 3

# 保存结果文件到本地
test['target'] = final_test
test[['id','dt','target']].to_csv('submit3.csv', index=None)

4 lightgbm 特征工程+回调函数

4.1 只加了特征工程

特征工程代码见3.2部分

4.2 再加个回调函数

回调函数参考了文章Datawhale AI夏令营 机器学习Task2 笔记打卡-CSDN博客的尝试3代码部分（72~92行）

from lightgbm.callback import log_evaluation
from lightgbm import early_stopping, log_evaluation

def time_model(lgb, train_df, test_df, cols):
    # 训练集和验证集切分
    trn_x, trn_y = train_df[train_df.dt>=31][cols], train_df[train_df.dt>=31]['target']
    val_x, val_y = train_df[train_df.dt<=30][cols], train_df[train_df.dt<=30]['target']
    # 构建模型输入数据
    train_matrix = lgb.Dataset(trn_x, label=trn_y)
    valid_matrix = lgb.Dataset(val_x, label=val_y)
   
    # 设置参数
    params = {
        'objective': 'regression',
        'metric': 'mse',
        'boosting_type': 'gbdt',
        'num_leaves': 31,
        'learning_rate': 0.05,
        'feature_fraction': 0.9
    }
    
    # 设置回调函数
    callbacks = [
        log_evaluation(period=100),  # 每100轮记录一次日志
        early_stopping(stopping_rounds=500)  # 500轮没有提升时提前停止
    ]
    
    # 训练模型
    model = lgb.train(params, train_matrix, num_boost_round=50000, valid_sets=[train_matrix, valid_matrix],
                    callbacks=callbacks)

    # 验证集和测试集结果预测
    val_pred = model.predict(val_x, num_iteration=model.best_iteration)
    test_pred = model.predict(test_df[cols], num_iteration=model.best_iteration)
    # 离线分数评估
    score = mean_squared_error(val_pred, val_y)
    print(score)
       
    return val_pred, test_pred
    
lgb_oof, lgb_test = time_model(lgb, train, test, train_cols)

# 保存结果文件到本地
test['target'] = lgb_test
test[['id','dt','target']].to_csv('submit2.csv', index=None)

5 更多挣扎：还在整，回头补

打算试试看更猛的特征工程，但是用方法2（将浮点型数据float64转换为float32或float16）解决内存爆炸。

记得点赞收藏关注