# let's begin by importing the basic libraries

import pandas as pd

import numpy as nb

import matplotlib.pyplot as plt

import csv

import seaborn as sns

%matplotlib inline

# enable seaborn styles

sns.set()

sns.set_style('whitegrid')

# read the .csv file and turn it into a normal Pandas Data Frame

cosmetics = pd.read_csv('/Users/janetliu/Desktop/cosmetics/cscpopendata-openrefinesample.csv')

# we can check the first 5 rows to see what our data looks like

cosmetics.head(5)

# i am going to do a quick search to make sure i have uploaded the sample data

print('Total reports: ' + str(len(cosmetics)))

Total reports: 1000

# the next step is to check our data type objects and see what our values are listed under

cosmetics.dtypes

CDPHId                      int64

ProductName                object

CSFId                     float64

CSF                        object

CompanyId                   int64

CompanyName                object

BrandName                  object

PrimaryCategoryId           int64

PrimaryCategory            object

SubCategoryId               int64

SubCategory                object

CasId                       int64

CasNumber                  object

ChemicalId                  int64

ChemicalName               object

InitialDateReported        object

MostRecentDateReported     object

DiscontinuedDate          float64

ChemicalCreatedAt          object

ChemicalUpdatedAt          object

ChemicalDateRemoved        object

ChemicalCount               int64

dtype: object

# most of the data we need are listed as objects

## lets change the values into a date time format

cosmetics['InitialDateReported'] = pd.to_datetime(cosmetics['InitialDateReported'])

cosmetics['MostRecentDateReported'] = pd.to_datetime(cosmetics['MostRecentDateReported'])

cosmetics['DiscontinuedDate'] = pd.to_datetime(cosmetics['DiscontinuedDate'])

cosmetics['MostRecentDateReported'] = pd.to_datetime(cosmetics['MostRecentDateReported'])

cosmetics['DiscontinuedDate'] = pd.to_datetime(cosmetics['DiscontinuedDate'])

cosmetics['ChemicalCreatedAt'] = pd.to_datetime(cosmetics['ChemicalCreatedAt'])

cosmetics['ChemicalUpdatedAt'] = pd.to_datetime(cosmetics['ChemicalUpdatedAt'])

cosmetics['ChemicalDateRemoved'] = pd.to_datetime(cosmetics['ChemicalDateRemoved'])

### and categories

cosmetics.dtypes

for col in [

    'CompanyName','BrandName', 'PrimaryCategory', 'SubCategory', 'ChemicalName']: cosmetics[col] = cosmetics[col].astype('category')

# then re-print the datatypes to make sure the changes were made

cosmetics.dtypes

CDPHId                             int64

ProductName                       object

CSFId                            float64

CSF                               object

CompanyId                          int64

CompanyName                     category

BrandName                       category

PrimaryCategoryId                  int64

PrimaryCategory                 category

SubCategoryId                      int64

SubCategory                     category

CasId                              int64

CasNumber                         object

ChemicalId                         int64

ChemicalName                    category

InitialDateReported       datetime64[ns]

MostRecentDateReported    datetime64[ns]

DiscontinuedDate          datetime64[ns]

ChemicalCreatedAt         datetime64[ns]

ChemicalUpdatedAt         datetime64[ns]

ChemicalDateRemoved       datetime64[ns]

ChemicalCount                      int64

dtype: object

# now that our values are in the right format, we can begin to answer the questions, 

###'Who are the top 10 cosmetic brands with the most reported chemicals in their products?' and 

#### 'What are the top 10 toxic chemicals most commonly found in beauty products?'

##### i hypothesize that high end, expensive brands will contain less toxic chemicals than cheaper, drugstore brands

###### let's perform a value_counts on BrandName to answer the first question

cosmetics.BrandName.value_counts()[:10].plot(kind='bar', color='palevioletred') # value_count to show the frequency, and bar for barplot 

plt.xticks(rotation=-45) # set the x label angle 

plt.title('Top 10 Cosmetic Brands with the Highest Number of Reported Chemicals') #give a title

plt.ylabel('Chemical Count') #and label names

plt.xlabel('Cosmetic Brands')

plt.rcParams['figure.figsize'] = figsize=(20,7)  # note that altering rcParams can alter the inline graph size

# the plot shows 'tarte', 'sephora' and 'DND' are the three top cosmetic brands with the most reported chemicals

## this is alarming as i own cosmetic products from 5 brands listed on here

### this plot also shows high end, expensive brands are equally likely to contain the same number of toxic chemicals

#### what's also disturbing is that there is only one nail polish brand (DND) in the top 10

##### while the rest are brands that sell face products such as innisfree, a brand i routinely use for their face masks

###### let's perform another value_counts on ChemicalName to answer our next question

cosmetics.ChemicalName.value_counts()[:10].plot(kind='bar', color='palevioletred')

plt.xticks(rotation=-90)

plt.title('Top 10 Most Common Chemicals found in Cosmetic Products')

plt.ylabel('Chemical Count')

plt.xlabel('Chemical Name')

plt.rcParams['figure.figsize'] = figsize=(20,7) 

# Of the 1,000 entries, the most common chemical found in cosmetic products is called 'Titanium dioxide'

## which can also be found in paints, plastics, and rubber

### 'taic' is also commonly found in rubber, and 'carbon black' is found in coal

#### i am also curious to know the frequency of the latest chemicals reported in 2020

### let's do this by creating a frequency timeline

cosmetics.MostRecentDateReported.value_counts().plot(kind='line',color='palevioletred') #set to linepot 

plt.title('Timeline of most recent Chemicals reported in 2020')

plt.ylabel('Number of reports')

plt.xlabel('Date')

plt.rcParams['figure.figsize'] = (20,7) 

# the graph shows there were over 175 reports made in late February, while the number of reports are declining

## let's move on to seaborn for our next visualizations

# for the second part, i want to find out, 'Which everyday cosmetic products include the most number of chemicals?'

## and 'How many chemicals can be found in a cosmetic product?'

### to answer the first question, let's create a count plot 

sns.catplot( #run a categorical plot with seaborn

    y="PrimaryCategory",

    kind="count", #use a countplot to find the frequency of reported chemicals

    height=7, #make the plot 7 units high

    palette='pastel',

    aspect=2, #height should be two times width

    data=cosmetics);

plt.title('Number of Chemicals found in Everyday Cosmetic Products')

plt.xlabel('Number of Chemicals')

plt.ylabel('Types of Cosmetic Products')

# the count plot shows 'make up products (non-permanent)' contains the most reported number of chemicals

## but it is unclear what 'make up products' include

### let's create another countplot to see the subcategories in this group

sns.catplot( #set up categorical scatterplots for each subcategory

    y="SubCategory", 

    col="PrimaryCategory",

    color='palevioletred',

    col_wrap=2, #set the number of charts per row

    data=cosmetics, 

    kind="count", 

    height=6, 

    aspect=1.5)

# showing the number of subcategories within each primary category gives me a sense of what kind of 

## every day beauty products are found in 'make up products (non-permanent)''

### these include lip colors, eye shadows, face powders, and blushes which are products i should be more mindful of

#### and read the ingredients when purchasing 

#### lastly, i want to create a violin chart to visualize the distribution of chemicals for each brand

##### i can do this by looking at the total number of current chemicals reported for each brand's products

sns.catplot(

    x="BrandName", 

    y="ChemicalCount", 

    kind="violin", 

    palette='pastel',

    dpi=300, inner="stick", 

    aspect=1.5, 

    data=cosmetics);

plt.title('Which beauty products contain the most chemicals?')

plt.xlabel('Cosmetic Brands')

plt.ylabel('Chemical Count in each Product')

plt.xticks(rotation=-90) #rotate x-labels

fig = plt.gcf()

fig.set_size_inches(20, 10)