

In [1]:

    
import os
import pickle
import numpy as np
import pandas as pd
from os import path
import seaborn as sns
import matplotlib.pyplot as plt
from dotenv import load_dotenv, find_dotenv
%matplotlib notebook

sns.set(color_codes=True)

RAW_DATA_DIR = os.environ.get("RAW_DATA_DIR")
RAW_DATA_DIR='D:\\Documents\\Notebooks\\CapstoneProject\\data\\raw'



In [2]:

    
data = pd.read_csv(path.join(RAW_DATA_DIR, 'gender_age_train.csv'))
data = data.sort_values(by='age')
labels = data['group']



In [3]:

    
f = {}
for l in labels:
    if l in f:
        f[l] += 1
    else:
        f[l] = 1

sex = {}
for g in data['gender']:
    if g in sex:
        sex[g] += 1
    else:
        sex[g] = 1

for s in sex:
    sex[s] = float(sex[s])/len(labels)
for e in f:
    f[e] = float(f[e])/len(labels)



In [20]:

    
print data['gender'].value_counts()
print 'Sexes frequency: {:.3f} males, {:.3} females'.format(sex['M'], sex['F'])









    



M    47904
F    26741
Name: gender, dtype: int64
Sexes frequency: 0.642 males, 0.358 females



In [32]:

    
ax = sns.countplot(x='gender', data=data)
ax.set_title('Males and Females instances')









    














    











    Out[32]:





<matplotlib.text.Text at 0xc105748>



In [34]:

    
plt.figure()
age = sns.distplot(data['age'], hist=False)
age.set_title('Age distribution')









    














    











    Out[34]:





<matplotlib.text.Text at 0xaf39128>



In [38]:

    
plt.figure()
sns.kdeplot( data['age'][ data['gender'] == 'F' ], label='Female')
sns.kdeplot( data['age'][ data['gender'] == 'M' ], label='Male').set_title(
                                        'Age distribution divided by gender')









    














    











    Out[38]:





<matplotlib.text.Text at 0xdb5bcf8>



In [42]:

    
plt.figure()
ax = sns.countplot(x='group', data=data[data.gender == 'M'], saturation=1)
ax.set_title('Male users in each group')









    














    











    Out[42]:





<matplotlib.text.Text at 0xb60ca90>



In [43]:

    
plt.figure()
ax = sns.countplot(x='group', data=data[data.gender == 'F'], saturation=1)
ax.axes.set_ylim(0,10000)
ax.set_title('Female users in each group')









    














    











    Out[43]:





<matplotlib.text.Text at 0xe0a4828>

These plots shows that the classes are unbalanced. Young people are largely more numerous than old, also there is a higher representation of males over females. Among the older, there are more data coming from males rather than females. This result is consistent with Chinese demographic and the fact that the use of technology is more diffused among younger generations. Therefore these data don't have to be resempled in order to create a uniform distribution of the classes. COMMENT ON DIFFERENT BIN WIDTH AGE-WISE



In [ ]: