数据集来自kaggle

import numpy as np
import pandas as pd

数据读取

dataset = pd.read_csv("./flavors_of_cacao.csv")

dataset.columns = dataset.columns.map(lambda x:x.replace("\n"," "))
dataset.columns = dataset.columns.map(lambda x:x.replace("\xa0",""))
dataset.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1795 entries, 0 to 1794
Data columns (total 9 columns):
Company (Maker-if known)            1795 non-null object
Specific Bean Origin or Bar Name    1795 non-null object
REF                                 1795 non-null int64
Review Date                         1795 non-null int64
Cocoa Percent                       1795 non-null object
Company Location                    1795 non-null object
Rating                              1795 non-null float64
Bean Type                           1794 non-null object
Broad Bean Origin                   1794 non-null object
dtypes: float64(1), int64(2), object(6)
memory usage: 126.3+ KB

每个列的含义如下：

• Company：生产公司
• Specific Bean Origin or Bar Name：产品名称
• REF：不祥
• Review Date：
• Cocoa Percent：可可含量
• Company Location：公司地址
• Rating：等级
• Bean Type：可可豆类型
• Broad Bean Origin：原产地

数据预处理

缺失值丢弃

dataset_nona = dataset.dropna()
dataset_nona = dataset_nona.drop(["REF"],axis=1)
dataset_nona.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1793 entries, 0 to 1794
Data columns (total 8 columns):
Company (Maker-if known)            1793 non-null object
Specific Bean Origin or Bar Name    1793 non-null object
Review Date                         1793 non-null int64
Cocoa Percent                       1793 non-null object
Company Location                    1793 non-null object
Rating                              1793 non-null float64
Bean Type                           1793 non-null object
Broad Bean Origin                   1793 non-null object
dtypes: float64(1), int64(1), object(6)
memory usage: 126.1+ KB

百分比转换

dataset_nona["Cocoa Percent"] = dataset_nona["Cocoa Percent"].map(lambda x:float(x.strip('%')) / 100)
dataset_nona.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1793 entries, 0 to 1794
Data columns (total 8 columns):
Company (Maker-if known)            1793 non-null object
Specific Bean Origin or Bar Name    1793 non-null object
Review Date                         1793 non-null int64
Cocoa Percent                       1793 non-null float64
Company Location                    1793 non-null object
Rating                              1793 non-null float64
Bean Type                           1793 non-null object
Broad Bean Origin                   1793 non-null object
dtypes: float64(2), int64(1), object(5)
memory usage: 126.1+ KB

问题分析

Where are the best cocoa beans grown?

best_been = dataset_nona[["Broad Bean Origin","Rating"]]

best_been_data = best_been.groupby(["Broad Bean Origin"]).apply(np.mean)
best_been_data.sort_values(by="Rating",inplace=True)
print(best_been_data[-10:])

                              Rating
Broad Bean Origin                   
Dominican Rep., Bali            3.75
Peru, Belize                    3.75
Ven.,Ecu.,Peru,Nic.             3.75
DR, Ecuador, Peru               3.75
Venez,Africa,Brasil,Peru,Mex    3.75
Dom. Rep., Madagascar           4.00
Venezuela, Java                 4.00
Gre., PNG, Haw., Haiti, Mad     4.00
Guat., D.R., Peru, Mad., PNG    4.00
Peru, Dom. Rep                  4.00

可看出最好的可可豆生长在秘鲁的Dom. Rep，危地马拉的D.R., Peru, Mad., PNG等地

Which countries produce the highest-rated bars?

best_country = dataset_nona[["Company Location","Rating"]]
best_country.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1793 entries, 0 to 1794
Data columns (total 2 columns):
Company Location    1793 non-null object
Rating              1793 non-null float64
dtypes: float64(1), object(1)
memory usage: 42.0+ KB

best_country_data = best_country.groupby(["Company Location"]).apply(np.mean)
best_country_data.sort_values(by=["Rating"],inplace=True)
print(best_country_data[-10:])

                    Rating
Company Location          
Guatemala         3.350000
Australia         3.357143
Poland            3.375000
Brazil            3.397059
Vietnam           3.409091
Iceland           3.416667
Philippines       3.500000
Netherlands       3.500000
Amsterdam         3.500000
Chile             3.750000

可以看出生产出巧克力较好的是智利，荷兰等地

what’s the relationship between cocoa solids percentage and rating?

best_coco = dataset_nona[["Cocoa Percent","Rating"]]
best_coco.columns = best_coco.columns.map(lambda x:x.replace(" ",""))
best_coco.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1793 entries, 0 to 1794
Data columns (total 2 columns):
CocoaPercent    1793 non-null float64
Rating          1793 non-null float64
dtypes: float64(2)
memory usage: 42.0 KB

print(best_coco.corr())

              CocoaPercent    Rating
CocoaPercent      1.000000 -0.164758
Rating           -0.164758  1.000000

import matplotlib.pyplot as plt
plt.close()
# print(best_coco["CocoaPercent"])
plt.scatter(best_coco["CocoaPercent"].values,best_coco["Rating"].values)
plt.show()

散点图

可以看出巧克力质量和含可可量没有明显的关系

探索分析

print(dataset_nona.groupby(["Review Date"]).apply(lambda x:x["Rating"].sum() / x.shape[0]))

Review Date
2006    3.125000
2007    3.162338
2008    2.994624
2009    3.073171
2010    3.148649
2011    3.251524
2012    3.181701
2013    3.197011
2014    3.189271
2015    3.246491
2016    3.226027
2017    3.312500
dtype: float64

coco_type = dataset_nona[["Bean Type","Rating"]]
coco_type = coco_type.groupby(["Bean Type"]).apply(np.mean)
print(coco_type.sort_values(by="Rating")[-10:])

                          Rating
Bean Type                       
Amazon, ICS                3.625
Criollo (Ocumare 77)       3.750
Trinitario, TCGA           3.750
Blend-Forastero,Criollo    3.750
Amazon mix                 3.750
Trinitario, Nacional       3.750
Forastero (Amelonado)      3.750
Trinitario (85% Criollo)   3.875
Criollo (Wild)             4.000
Criollo (Ocumare 67)       4.000

最好的可可豆是Criollo