In [1]:
import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
from matplotlib import pyplot as plt
import os
print(os.listdir("../input"))
# Any results you write to the current directory are saved as output.
In [2]:
df = pd.read_csv("../input/googleplaystore.csv")
df.head()
Out[2]:
In [3]:
features = ['Category','Rating','Reviews','Size','Type','Price','Content Rating']
y = ['Installs']
dropping a row with app name 'Life Made WI-Fi Touchscreen Photo Frame' as it has shifted data:
In [4]:
print(df.columns)
df = df.set_index("App")
print(df.columns)
df = df.drop("Life Made WI-Fi Touchscreen Photo Frame")
df = df.reset_index()
print(df.columns)
#df[df['App'] == 'Life Made WI-Fi Touchscreen Photo Frame']
Now, we impute the median rating for all the NaN values in the column Rating 1.Remove NANs from type and other features. 2.write a function and change the value of type (free=0 and paid = 1) 3.plot individual graphs
In [5]:
df['Rating'].fillna(df.median()['Rating'],inplace=True)
Function for converting size into a uniform unit(M). Varies with device => -1
In [6]:
#print(df['Size'])
def size_convert(d):
s=d[4]
if s[len(s)-1] == "M":
return float(s[:len(s)-1])
elif s[len(s)-1] == "k" or s[len(s)-1] == "K":
return (float(s[0:len(s)-1])/1000)
else:
return -1
In [7]:
df['Size'] = df.apply(size_convert, axis=1)
In [8]:
def clean_Installs(b):
a=b[5]
return(a.replace("+","").replace(",",""))
df['Installs'] = df.apply(clean_Installs, axis=1)
In [9]:
df.head(20)
Out[9]:
In [10]:
def clean_Price(b):
a=b[7]
return(a.replace("$",""))
df['Price'] = df.apply(clean_Price, axis=1)
In [11]:
df.Reviews = df.Reviews.apply(int)
df.Installs = df.Installs.apply(int)
df.Price = df.Price.apply(float)
Making the single record with Type as NaN 'Free'
In [12]:
fil1 = df.Type.isnull()
fil2 = df.Price == 0
df['Type'].fillna("Free",inplace=True)
Making dummies for the types: Free and Paid
In [13]:
#dummy = pd.get_dummies(df['Type'])
#df = pd.concat([df,dummy],axis = 1)
#df.head()
In [14]:
def make_others(d):
a = d[1]
k=df.Category.unique()
k=k.tolist()
k.remove('SPORTS')
k.remove('COMMUNICATION')
k.remove('ENTERTAINMENT')
k.remove('HEALTH_AND_FITNESS')
k.remove('GAME')
k.remove('FAMILY')
k.remove('SOCIAL')
k.remove('PHOTOGRAPHY')
k.remove('TRAVEL_AND_LOCAL')
k.remove('TOOLS')
k.remove('PRODUCTIVITY')
k.remove('VIDEO_PLAYERS')
k.remove('NEWS_AND_MAGAZINES')
k.remove('LIFESTYLE')
k.remove('PERSONALIZATION')
k.remove('LIBRARIES_AND_DEMO')
k.remove('SHOPPING')
if a in k:
return "OTHERS"
else:
return a
In [15]:
df['Category'] = df.apply(make_others, axis=1)
In [16]:
df_cat = df.groupby(df['Category']).agg({'Installs':np.sum}).reset_index()
plt.bar(df_cat.Category,df_cat.Installs)
plt.grid()
plt.xticks(rotation=90)
plt.xlabel('Category')
plt.ylabel('Installs(*10^10)')
plt.title('Category graph')
plt.show()
In [17]:
df_rat = df.groupby(df['Rating']).agg({'Installs':np.average}).reset_index()
plt.plot(df_rat.Rating,df_rat.Installs,marker='o')
plt.grid()
#plt.xticks(rotation=90)
plt.xlabel('Rating')
plt.ylabel('Installs(*10^10)')
plt.title('Rating graph')
plt.show()
In [18]:
print(df.Reviews.max(),df.Reviews.min())
import numpy as np
custom_bucket_array = np.arange(0,78158306,10000000)
df["Reviews"] = pd.cut(df["Reviews"],custom_bucket_array)
fin1 = df.groupby(df.Reviews).agg({'Installs':np.average}).reset_index()
#print(fin1)
categories = fin1['Reviews'].cat.categories
ind = np.array([x for x, _ in enumerate(categories)])
width = 0.15
In [19]:
plt.bar(ind,fin1["Installs"],width, label='Installs')
plt.xticks(ind, categories)
plt.xticks(rotation = 90)
plt.xlabel('Rating')
plt.ylabel('Installs(*10^10)')
plt.title('Rating graph')
plt.show()
In [20]:
from scipy import stats
df_1 = df[df['Size'] != -1]
df_size = df_1.groupby(df_1['Size']).agg({'Installs':np.sum}).reset_index()
plt.bar(df_size.Size,df_size.Installs)
plt.grid()
#plt.xticks(rotation=90)
plt.xlabel('Size')
plt.ylabel('Installs(*10^9)')
plt.title('Size graph')
plt.show()
In [21]:
df_type = df.groupby(df['Type']).agg({'Installs':np.average}).reset_index()
plt.bar(df_type.Type,df_type.Installs)
plt.grid()
#plt.xticks(rotation=90)
plt.xlabel('Type')
plt.ylabel('Installs(*10^9)')
plt.title('Type graph')
plt.show()
In [22]:
plt.scatter(df.Price,df.Installs)
plt.grid()
#plt.xticks(rotation=90)
plt.xlabel('Price')
plt.ylabel('Installs(*10^9)')
plt.title('Price graph')
plt.show()
In [23]:
df = df[df['Price']<290]
In [24]:
df_2 = df[df['Price']!=0]
df_type = df_2.groupby(df_2['Price']).agg({'Installs':np.sum}).reset_index()
plt.scatter(df_type.Price,df_type.Installs)
plt.grid()
#plt.xticks(rotation=90)
plt.xlabel('Price')
plt.ylabel('Installs(*10^7)')
plt.title('Price graph')
plt.show()
In [25]:
df_cnt = df.groupby(df['Content Rating']).agg({'Installs':np.average}).reset_index()
plt.bar(df_cnt['Content Rating'],df_cnt.Installs)
plt.grid()
#plt.xticks(rotation=90)
plt.xlabel('Content Rating')
plt.ylabel('Installs(*10^11)')
plt.xticks(rotation=90)
plt.title('Content Rating graph')
plt.show()
In [ ]: