Introduction¶
For this investigation, I chose to look into the TMDb movie data set. This data set contains movie information such as cast and genre, and also financial information regarding the movie's budget and revenue. It also includes adjusted budget and revenue values in terms of 2010 dollars, accounting for inflation over time. These columns are called 'budget_adj' and 'revenue_adj' respectively.
The questions that I plan to explore will revolve around traits that relate to the revenue of the movies, particulary the adjusted revenue in terms of 2010 dollars since movies across multiple release years will be explored.
In [1]:
import pandas as pd
import matplotlib.pyplot as plt
% matplotlib inline
# Change floats from displaying in scientific notation
pd.options.display.float_format = '{:20,.2f}'.format
In [2]:
df = pd.read_csv('tmdb-movies.csv')
df.head()
Out[2]:
Here's a look at the statistical overviews:¶
In [3]:
df.describe()
Out[3]:
Here's a look at the column data types and counts:¶
In [4]:
df.info()
Data Cleaning¶
First let's check for duplicates.¶
In [5]:
df[df.duplicated() == True].shape
Out[5]:
We have one duplicate. The following cell will remove it.¶
In [6]:
df.drop_duplicates(inplace=True)
Check the duplicate was removed.¶
In [7]:
df[df.duplicated() == True].shape
Out[7]:
One of the columns of interest is 'genres', and there are some movies for which this data is missing.¶
In [8]:
df[df['genres'].isnull()].shape
Out[8]:
These rows will be removed in the following cell.¶
In [9]:
df = df.dropna(subset = ['genres'])
Check that these rows were removed.¶
In [10]:
df[df['genres'].isnull()].shape
Out[10]:
There are also rows which have a value of 0 for either the 'budget_adj' or the 'revenue_adj' columns.¶
In [11]:
df.query("(budget_adj == 0) | (revenue_adj == 0)").shape
Out[11]:
The next cell will remove those rows.¶
In [12]:
df = df[df['budget_adj'] > 0]
df = df[df['revenue_adj'] > 0]
Check that these rows were removed.¶
In [13]:
df.query("(budget_adj == 0) | (revenue_adj == 0)").shape
Out[13]:
Here's a plot showing the relationship between 'budget_adj' and 'revenue_adj':¶
In [14]:
ax = df.plot(x='budget_adj', y='revenue_adj', kind='scatter', title='Adjusted Budget and Adjusted Revenue',figsize=(16,8))
ax.set_xticklabels(['{:,}'.format(int(x)) for x in ax.get_xticks().tolist()])
ax.set_yticklabels(['{:,}'.format(int(x)) for x in ax.get_yticks().tolist()]);
There is a lot of density of values when the 'budget_adj' is less than 100,000,000. The following 2 plots will separate the visuals at this value to get a closer view.¶
Plot for 'budget_adj' <= 100,000,000:¶
In [15]:
ax = df[df['budget_adj'] <= 100000000].plot(x='budget_adj', y='revenue_adj', kind='scatter', title='Adjusted Budget and Adjusted Revenue (Adjusted Budget <= 100,000,000)', figsize=(16,8))
ax.set_xticklabels(['{:,}'.format(int(x)) for x in ax.get_xticks().tolist()])
ax.set_yticklabels(['{:,}'.format(int(x)) for x in ax.get_yticks().tolist()]);
Plot for 'budget_adj' > 100,000,000:¶
In [16]:
ax = df[df['budget_adj'] > 100000000].plot(x='budget_adj', y='revenue_adj', kind='scatter', title='Adjusted Budget and Adjusted Revenue (Adjusted Budget > 100,000,000)', figsize=(16,8))
ax.set_xticklabels(['{:,}'.format(int(x)) for x in ax.get_xticks().tolist()])
ax.set_yticklabels(['{:,}'.format(int(x)) for x in ax.get_yticks().tolist()]);
Research Question 2: How does the movie's genre relate to revenue?¶
Specifically, are there certain genres that tend to generate more revenue than others?
Since a movie can have multiple genres, let's first get all the unique genre names.¶
In [17]:
genres_set = set()
for genres in df['genres'].iteritems():
for genre in genres[1].split('|'):
genres_set.add(genre)
genres_set = sorted(genres_set)
genres_set
Out[17]:
Now let's add a new column for each genre, with a value of True or False depending on if the movie is categorized as that specific genre.¶
In [18]:
for genre in genres_set:
df[genre] = df['genres'].str.contains(genre)
# Adjust the column slice so we can see some of the added individual genre columns
df.iloc[:, 13:].head()
Out[18]:
In [19]:
# This dictionary will be used to create a new data frame
genre_rev_adj_means_dict = {
'mean_revenue_adj': {}
}
# For each genre and associated mean adjusted revenue, add a key/value to the 'mean_revenue_adj' column
for genre in genres_set:
mean_rev_adj = df[df[genre]]['revenue_adj'].mean()
genre_rev_adj_means_dict['mean_revenue_adj'][genre] = mean_rev_adj
# Create and sort the data frame
df_mean_rev_adj_genre = pd.DataFrame.from_dict(genre_rev_adj_means_dict)
df_mean_rev_adj_genre.sort_values(by=['mean_revenue_adj'], inplace=True)
df_mean_rev_adj_genre
Out[19]:
Here's a plot of the data:¶
In [20]:
ax = df_mean_rev_adj_genre.plot(title='Average Adjusted Revenue By Genre', kind='bar', figsize=(16,8), legend=False)
ax.set_yticklabels(['{:,}'.format(int(x)) for x in ax.get_yticks().tolist()]);
Research Question 3: How has movie revenue changed over time?¶
Specifically, is there a trend of revenue increasing or decreasing over time?
Let's first get all the unique release years.¶
In [21]:
release_year_set = sorted(set(df['release_year']))
release_year_set
Out[21]:
In [22]:
# This dictionary will be used to create a new data frame
release_year_rev_adj_means_dict = {
'mean_revenue_adj': {}
}
# For each release year and associated mean adjusted revenue, add a key/value to the 'mean_revenue_adj' column
for year in release_year_set:
mean_rev_adj = df[df['release_year'] == year]['revenue_adj'].mean()
release_year_rev_adj_means_dict['mean_revenue_adj'][year] = mean_rev_adj
# Create the data frame
df_mean_rev_adj_release_year = pd.DataFrame.from_dict(release_year_rev_adj_means_dict)
df_mean_rev_adj_release_year
Out[22]:
Here's a plot of the data:¶
In [23]:
ax = df_mean_rev_adj_release_year.plot(title='Average Adjusted Revenue By Release Year', kind='bar', figsize=(16,8), legend=False)
ax.set_yticklabels(['{:,}'.format(int(x)) for x in ax.get_yticks().tolist()]);
Conclusions¶
Research Question 1: How does the movie's budget relate to revenue?¶
- There does not appear to be any correlation between budget and revenue.
Research Question 2: How does the movie's genre relate to revenue?¶
- There does appear to be a trend that some movie genres tend to generate a higher revenue, with the Animation and Adventure genres being the top revenue generators.
- Foreign and Documentary genres show as being the lowest revenue generators.
Research Question 3: How has movie revenue changed over time?¶
- Since the 1960s revenues have fluctuated, with the 1960s and 1970s being the period of having the highest revenues.
- From the 1980s and later there have been some relatively small fluctuations but overall have been relatively steady.
Limitations Of The Data Set¶
- The TMDb data set has been used for the analysis and worked with traits such as popularity, revenue and runtime. Analysis is limited to only the provided data. For example, the data set does not confirm that the director of every movie is listed.
- There is no normalization or exchanges rate or currency conversion considered during this analysis, and is limited to the numerical values of revenue.
- Dropping missing or Null values from variables of our interest might skew our analysis and could show unintentional bias towards relationships being analyzed. This could also cause increased disparities of representation for certain groups of data, in addition to any already present in the data set.
- It should be noted that a movie can represent multiple genres. In the genre analysis, a movie is taken into account for each genre it represents.