ium_434684/ium_zadanie1.py

import pandas as pd
import seaborn as sns
import datetime
import matplotlib.pyplot as plt

netflix=pd.read_csv('netflix_titles_enriched.csv')

netflix_cleaned = netflix[netflix.rottentomatoes_audience_score > 0].sort_values(by = 'rottentomatoes_audience_score')
netflix_cleaned.rottentomatoes_audience_score /= 100
netflix_cleaned.drop(['rottentomatoes_audience_#reviews',
       'rottentomatoes_audience_review', 'rottentomatoes_tomatometer_score',
       'rottentomatoes_critics_#reviews', 'rottentomatoes_critic_review'], axis = 1)

netflix_cleaned.date_added = netflix_cleaned.date_added.dropna().apply(lambda x: datetime.datetime.strptime(x[1:] if x[0] == ' ' else x, '%B %d, %Y'))

netflix_cleaned.update(netflix_cleaned.select_dtypes(include = 'object').apply(lambda col: col.str.lower()))



movies = netflix_cleaned[netflix_cleaned.type == 'movie']
series = netflix_cleaned[netflix_cleaned.type == 'tv show']


movies.duration = movies.duration.str.extract(r'(\d*)( min)')[0].astype('int32')

from sklearn.preprocessing import MultiLabelBinarizer
mlb = MultiLabelBinarizer()

movies = movies.join(pd.DataFrame(mlb.fit_transform(movies.pop('listed_in').str.split(', ')),
                          columns=mlb.classes_,
                          index=movies.index))
movies.drop(['movies'], axis = 1)

movies = movies[['release_year', 'duration',
       'rottentomatoes_audience_score',
       'action & adventure', 'anime features', 'children & family movies',
       'classic movies', 'comedies', 'cult movies', 'documentaries', 'dramas',
       'faith & spirituality', 'horror movies', 'independent movies',
       'international movies', 'lgbtq movies', 'movies', 'music & musicals',
       'romantic movies', 'sci-fi & fantasy', 'sports movies',
       'stand-up comedy', 'thrillers']]

import sklearn
from sklearn.model_selection import train_test_split
movies_train, movies_test = sklearn.model_selection.train_test_split(movies,test_size=0.20, random_state=42)
movies_test, movies_val = sklearn.model_selection.train_test_split(movies_test,test_size=0.50, random_state=42)

movies_subsets = [movies_train, movies_test, movies_val]
for subset in movies_subsets:
    print(subset.shape[0])
    print(subset.describe(include='all'))

movies_subsets = pd.concat(movies_subsets, keys = ['Train','Test','Validation'])
movies_subsets = movies_subsets.reset_index()
ax = sns.boxplot(data = movies_subsets, x = 'level_0', y = 'rottentomatoes_audience_score')
ax.set(title = 'Audience score distribution between subsets', ylabel = 'Audience score on Rotten Tomatoes', xlabel = 'SUBSET')
#plt.show(ax)


series.duration = series.duration.str.extract(r'(\d*)( seasons?)')[0].astype('int32')

series = series.rename(columns = {'Unnamed: 0': 'Season'})
series['Id'] = series.Season.str.extract(r'(s\d+)(|\',\ )(\d+)')[0]
series.Season = series.Season.str.extract(r'(s\d+)(|\',\ )(\d+)')[2].astype('int32')
series = series[series.Season > 0]

mlb = MultiLabelBinarizer()

series = series.join(pd.DataFrame(mlb.fit_transform(series.pop('listed_in').str.split(', ')),
                          columns=mlb.classes_,
                          index=series.index))

series_train, series_test = sklearn.model_selection.train_test_split(series,test_size=0.20, random_state=42)
series_test, series_val = sklearn.model_selection.train_test_split(series_test,test_size=0.50, random_state=42)

series_subsets = [series_train, series_test, series_val]
for subset in series_subsets:
    print(subset.shape[0])
    print(subset.describe(include='all'))

series_subsets = pd.concat(series_subsets, keys = ['Train','Test','Validation'])
series_subsets = series_subsets.reset_index()
ax = sns.boxplot(data = series_subsets, x = 'level_0', y = 'rottentomatoes_audience_score')
ax.set(title = 'Audience score distribution between subsets', ylabel = 'Audience score on Rotten Tomatoes', xlabel = 'SUBSET')
#plt.show(ax)



movies_train.to_csv('movies_train.csv')
movies_test.to_csv('movies_test.csv')
movies_val.to_csv('movies_val.csv')