Sign In


The objective of this project is to deliver insights to understand customer demands better and thus help developers to popularize the product. The dataset is chosen from Kaggle. It is of 10k Play Store apps for analyzing the Android market. It consists of in total of 10841 rows and 13 columns.

In [1]:
!pip install numpy pandas seaborn matplotlib -q
In [2]:
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

%matplotlib inline
In [3]:
!pip install jovian opendatasets --upgrade --quiet
In [4]:
dataset_url = ''
In [5]:
import opendatasets as od
Please provide your Kaggle credentials to download this dataset. Learn more: Your Kaggle username: ritikars Your Kaggle Key: ········
0%| | 0.00/1.94M [00:00<?, ?B/s]
Downloading to .\google-play-store-apps
100%|█████████████████████████████████████████████████████████████████████████████| 1.94M/1.94M [00:00<00:00, 4.69MB/s]
In [6]:
google_playstore_df = pd.read_csv('google-play-store-apps/googleplaystore.csv')
Checking what columns are there in the dataset
In [7]:
Index(['App', 'Category', 'Rating', 'Reviews', 'Size', 'Installs', 'Type',
       'Price', 'Content Rating', 'Genres', 'Last Updated', 'Current Ver',
       'Android Ver'],
Checking Number of columns and rows
In [8]:
(10841, 13)
In [9]:
<class 'pandas.core.frame.DataFrame'> RangeIndex: 10841 entries, 0 to 10840 Data columns (total 13 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 App 10841 non-null object 1 Category 10841 non-null object 2 Rating 9367 non-null float64 3 Reviews 10841 non-null object 4 Size 10841 non-null object 5 Installs 10841 non-null object 6 Type 10840 non-null object 7 Price 10841 non-null object 8 Content Rating 10840 non-null object 9 Genres 10841 non-null object 10 Last Updated 10841 non-null object 11 Current Ver 10833 non-null object 12 Android Ver 10838 non-null object dtypes: float64(1), object(12) memory usage: 1.1+ MB
Descriptive statistics
In [10]:
In [11]:
import jovian
In [ ]:
In [12]:
App               Life Made WI-Fi Touchscreen Photo Frame
Category                                              1.9
Rating                                                 19
Reviews                                              3.0M
Size                                               1,000+
Installs                                             Free
Type                                                    0
Price                                            Everyone
Content Rating                                        NaN
Genres                                  February 11, 2018
Last Updated                                       1.0.19
Current Ver                                    4.0 and up
Android Ver                                           NaN
Name: 10472, dtype: object
Removing 10472 due to data mismatch in the column
In [13]:
# Row 10472 removed due to missing value of Category
google_playstore_df.drop(google_playstore_df.index[10472], inplace=True)
Removing '+' from the values of number of installs and converting it to numeric
In [14]:
google_playstore_df['Installs'] = google_playstore_df['Installs'].map(lambda x: x.rstrip('+'))
In [15]:
google_playstore_df['Installs'] = pd.to_numeric(google_playstore_df['Installs'].str.replace(',',''))
Removing '$' from the values of Price and converting it to numeric
In [16]:
google_playstore_df['Price'] = pd.to_numeric(google_playstore_df['Price'].str.replace('$',''))
In [17]:
<class 'pandas.core.frame.DataFrame'> Int64Index: 10840 entries, 0 to 10840 Data columns (total 13 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 App 10840 non-null object 1 Category 10840 non-null object 2 Rating 9366 non-null float64 3 Reviews 10840 non-null object 4 Size 10840 non-null object 5 Installs 10840 non-null int64 6 Type 10839 non-null object 7 Price 10840 non-null float64 8 Content Rating 10840 non-null object 9 Genres 10840 non-null object 10 Last Updated 10840 non-null object 11 Current Ver 10832 non-null object 12 Android Ver 10838 non-null object dtypes: float64(2), int64(1), object(10) memory usage: 1.2+ MB
In [18]:
(0, 1000000000)
As we can notice,there is a high variance in the number of installs. To remove this we are adding a new column to dataframe, which is the log of number of installs
In [19]:
google_playstore_df['log_installs'] = np.log2(google_playstore_df['Installs'])
C:\Users\SinghRit\Anaconda3\lib\site-packages\pandas\core\ RuntimeWarning: divide by zero encountered in log2 result = getattr(ufunc, method)(*inputs, **kwargs)
Checking, if the dataframe contains any duplicate values for the same App name
In [20]:
boolean = google_playstore_df['App'].duplicated().any()
In [21]:
ROBLOX                                                9
CBS Sports App - Scores, News, Stats & Watch Live     8
Candy Crush Saga                                      7
ESPN                                                  7
Duolingo: Learn Languages Free                        7
Dots & Co: A Puzzle Adventure                         1
Messenger L SMS, MMS                                  1
Super P Launcher for Android P 9.0 launcher, theme    1
Get Over Your Ex                                      1
DV Statistics                                         1
Name: App, Length: 9659, dtype: int64
As we can see from the above mentioned data, some of the apps are having multiple rows. Let's check out if there data is identical or not
In [22]:
As we can see from the above dataframe, ROBLOX app is having identical rows with difference in number of reviews. It may have happened that for the same app, the data has been scraped in different points of time. So we have kept row of an app with maximum number of reviews, assuming it to be the latest one.
In [23]:
Convertingtype of Reviews column to number
In [24]:
google_playstore_df['Reviews'] = pd.to_numeric(google_playstore_df['Reviews'].str.replace('$',''))
In [25]:
google_playstore_df = google_playstore_df.loc[google_playstore_df.groupby(['App'])['Reviews'].idxmax()]
Setting up the parameters for the plots
In [26]:
plt.rcParams['font.size'] = 14
plt.rcParams['figure.figsize'] = (9, 5)
plt.rcParams['figure.facecolor'] = '#00000000'

What are the top 20 apps present in the google play store as per their Genres?

In [27]:
top_genres = google_playstore_df.Genres.value_counts().reset_index().rename(columns={'Genres':'Count','index':'Genres'})
In [28]:
genres_installs = google_playstore_df.groupby(['Genres'])[['Installs']].sum()
In [29]:
top_genres_installs = pd.merge(top_genres, genres_installs, on='Genres')
top_20_genres_installs = top_genres_installs.head(20)
In [30]:
plt.ylabel("Number of application")
plt.title("Top 20 Genres")
sns.barplot(top_20_genres_installs.Genres, top_20_genres_installs.Count)
<matplotlib.axes._subplots.AxesSubplot at 0x219ca024ec8>
Which are the Genres that are getting installed the most in top 20 Genres?
In [31]:
plt.title("Installs according to Genres")
sns.barplot(top_20_genres_installs.Genres, top_20_genres_installs.Installs)
<matplotlib.axes._subplots.AxesSubplot at 0x219ca82ed08>
As we can see from the above two plots: Maxinum number of apps present in google play store comes under Tools, Entertainment and Education Genres but as per the installation and requirement in the market plot, scenario is not the same. Maximum installed apps comes under Communication, Tools and Productivity Genres.

Average rating

In [32]:
genres_ratings_df = google_playstore_df.groupby(['Genres'])[['Rating']].mean()
In [33]:
genres_installs_ratings = pd.merge(top_genres_installs, genres_ratings_df, on='Genres')

Highest and lowest rated Genres

In [34]:
genres_installs_ratings.sort_values('Rating', ascending =False, inplace=True)
In [35]:
highest_rated_genres = genres_installs_ratings.iloc[0:20]
lowest_rated_genres = genres_installs_ratings.iloc[-20:]
lowest_rated_genres = lowest_rated_genres[lowest_rated_genres['Rating'].notnull()]
In [36]:
plt.title("Ratings according to Genres")
sns.barplot(highest_rated_genres.Genres, highest_rated_genres.Rating)
<matplotlib.axes._subplots.AxesSubplot at 0x219ca928f48>
In [37]:
plt.title("Ratings according to Genres")
sns.barplot(lowest_rated_genres.Genres, lowest_rated_genres.Rating)
<matplotlib.axes._subplots.AxesSubplot at 0x219caa088c8>

What are the count of applications in each category differentiated by their type?

In [39]:
app_count = google_playstore_df.groupby(['Category','Type'])[['App']].count().reset_index().rename(columns={'App':'Count','index':'App'})
In [40]:
free_apps = app_count[app_count['Type']=='Free']
In [41]:
paid_apps = app_count[app_count['Type']=='Paid']
In [42]:
plt.legend("Count of applications in each category differentiated by their type")['Category'], free_apps['Count'])['Category'], paid_apps['Count']);
It looks like certain app categories have more free apps available for download than others. In our dataset, the majority of apps in Family, Food & Drink and Tools, as well as Social categories were free to install. At the same time Family, Sports, Tools and Medical categories had the biggest number of paid apps available for download.

How many apps were installed according to its type?

In [43]:
google_playstore_df['Gaming Category App'] = google_playstore_df['Category']=='GAME'
In [44]:
categoty_type_installs = google_playstore_df.groupby(['Category','Type'])[['Installs']].sum().reset_index()
In [46]:
categoty_type_installs['log_Installs'] = np.log2(categoty_type_installs['Installs'])
In [47]:
plt.title("Number of installs type wise according to Category")
sns.barplot('Category', 'log_Installs', hue='Type', data=categoty_type_installs);
It can be concludd that the number of free applications installed by the user are high when compared with the paid ones.

Size of the applications present in the datset are in MB and KB. Therefore for ease in data processing, converting entire size column to MB.

In [48]:
# Converting KB to MB 
google_playstore_df['Size'] = google_playstore_df['Size'].apply(lambda x: str(x).replace('Varies with device', 'NaN') if 'Varies with device' in str(x) else x)
google_playstore_df['Size'] = google_playstore_df['Size'].apply(lambda x: str(x).replace('M', '') if 'M' in str(x) else x)
google_playstore_df['Size'] = google_playstore_df['Size'].apply(lambda x: str(x).replace(',', '') if 'M' in str(x) else x)
google_playstore_df['Size'] = google_playstore_df['Size'].apply(lambda x: float(str(x).replace('k', '')) / 1000 if 'k' in str(x) else x)
google_playstore_df['Size'] = google_playstore_df['Size'].apply(lambda x: float(x))

How does size impact on the number of installs of any application?

In [49]:
sns.scatterplot(google_playstore_df['Size'], google_playstore_df['log_installs'], hue=google_playstore_df['Type'])
<matplotlib.axes._subplots.AxesSubplot at 0x219cb6decc8>
It is clear from the above mentioned plot that size may impact the number of installations. Bulky applications are less installed by the user.

Histogram of log installs

In [50]:
In [51]:
plt.xlabel("Log of Installs")
plt.title("Distribution of Logrithm of Installs")

Histogram of size

In [52]:
# to remove null values from size column
In [53]:
plt.title("Distribution of Size")
From the above histogram, it can be concluded that maximum number of applications present in the dataset are of small size.
In [54]:
user_reviews_df = pd.read_csv('google-play-store-apps/googleplaystore_user_reviews.csv')

Merging dataframes on apps

In [56]:
merged_df = google_playstore_df.merge(user_reviews_df, on="App")
In the merged dataframe, we have three new columns i.e.. Sentiment, Sentiment Polarity and Sentiment Subjectivity. Sentiment basically determines the attitude or the emotion of the writer, i.e., whether it is positive or negative or neutral. Sentiment Polarity is float which lies in the range of [-1,1] where 1 means positive statement and -1 means a negative statement. Sentiment Subjectivity generally refer to personal opinion, emotion or judgment, which lies in the range of [0,1].

What is the distribution of type of reviews in the dataset?

In [57]:
category_sentiment = merged_df.groupby(['Category','Sentiment']).size().reset_index(name='Sentiment Count')
In [58]:
category_sentiment['log_sentiment_count'] = np.log2(category_sentiment['Sentiment Count'])
In [59]:
plt.title("Number of installs type wise according to Genres")
sns.barplot('Category', 'log_sentiment_count', hue='Sentiment', data=category_sentiment);

Histogram of subjectivity

In [60]:
plt.title("Distribution of Subjectivity")
It can be seen that maximum number of sentiment subjectivity lies between 0.4 to 0.7. From this we can conclude that macimum number of users give reviews to the applications, according to their experience.

Does sentiment_subjectivity proportional to sentiment_polarity?

In [61]:
sentimet_subjectivity_polarity = merged_df.groupby(['Category'])['Sentiment_Polarity','Sentiment_Subjectivity'].mean().reset_index().rename(columns={'Sentiment_Polarity':'Sentiment_polarity_avg','Sentiment_Subjectivity':'sentiment_subjectivity_avg'})

C:\Users\SinghRit\Anaconda3\lib\site-packages\ FutureWarning: Indexing with multiple keys (implicitly converted to a tuple of keys) will be deprecated, use a list instead. """Entry point for launching an IPython kernel.
In [62]:
merged_df['Sentiment_Subjectivity'] = merged_df['Sentiment_Subjectivity'].abs()
In [63]:
merged_df['Sentiment_Polarity'] = merged_df['Sentiment_Polarity'].abs()
In [64]:
sns.scatterplot(merged_df['Sentiment_Subjectivity'], merged_df['Sentiment_Polarity'])
plt.title("Does sentiment_subjectivity proportional to sentiment_polarity")
Text(0.5, 1.0, 'Does sentiment_subjectivity proportional to sentiment_polarity')