KaggleFutureSales“”竞赛XGB_model_final

介绍
“未来销售 "Future Sales"”竞赛是HSE莫斯科大学“高级机器学习”专业“如何赢得数据科学”课程的最终评估。目标是根据历史数据预测特定商店中商品的每月销售额。销售数量在 0 到 20 之间。

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.metrics import mean_squared_error
from sklearn.model_selection import train_test_split
from sklearn.base import clone
from sklearn.model_selection import GridSearchCV
sns.set(style&＃61;"darkgrid")import os
for dirname, _, filenames in os.walk(&＃39;/kaggle/input&＃39;):for filename in filenames:print(os.path.join(dirname, filename))

# load data
items&＃61;pd.read_csv("/kaggle/input/competitive-data-science-predict-future-sales/items.csv")
shops&＃61;pd.read_csv("/kaggle/input/competitive-data-science-predict-future-sales/shops.csv")
cats&＃61;pd.read_csv("/kaggle/input/competitive-data-science-predict-future-sales/item_categories.csv")
train&＃61;pd.read_csv("/kaggle/input/competitive-data-science-predict-future-sales/sales_train.csv")
test&＃61;pd.read_csv("/kaggle/input/competitive-data-science-predict-future-sales/test.csv")

数据清洗¶

去除原始数据清理异常值并添加变量

print(f&＃39;items.csv : {items.shape}&＃39;)
items.info()

 Remove Nan

train.isnull().sum()

 去除异常值

plt.figure(figsize&＃61;(10, 4))
plt.xlim(-100, 3000)
flierprops &＃61; dict(marker&＃61;&＃39;o&＃39;, markerfacecolor&＃61;&＃39;purple&＃39;, markersize&＃61;6,linestyle&＃61;&＃39;none&＃39;, markeredgecolor&＃61;&＃39;black&＃39;)
sns.boxplot(x&＃61;train.item_cnt_day, flierprops&＃61;flierprops)plt.figure(figsize&＃61;(10, 4))
plt.xlim(train.item_price.min(), train.item_price.max()*1.1)
sns.boxplot(x&＃61;train.item_price, flierprops&＃61;flierprops)

 我们将从数据集中删除明显的异常值 - 一天销售超过 1,000 件的商品和价格超过 300,000 件的商品。

train &＃61; train[(train.item_price <300000) & (train.item_cnt_day <1000)]

 从train中删除项目价格为负的所有行。 对于可以退款的售出商品数量&＃xff0c;我们会将小于 1 的值设为零以消除负值。

train &＃61; train[train.item_price > 0].reset_index(drop&＃61;True)
train.loc[train.item_cnt_day <1, "item_cnt_day"] &＃61; 0