deansublett

# Testing our content-based recommendation system with the seminal film Spy Kids
give_rec('Spy Kids')

deansublett

from sklearn.metrics.pairwise import sigmoid_kernel

# Compute the sigmoid kernel
sig = sigmoid_kernel(tfv_matrix, tfv_matrix)

# Reverse mapping of indices and movie titles
indices = pd.Series(movies_clean.index, index=movies_clean['original_title']).drop_duplicates()

# Credit to Ibtesam Ahmed for the skeleton code
def give_rec(title, sig=sig):

deansublett

from sklearn.feature_extraction.text import TfidfVectorizer

# Using Abhishek Thakur's arguments for TF-IDF
tfv = TfidfVectorizer(min_df=3,  max_features=None, 
            strip_accents='unicode', analyzer='word',token_pattern=r'\w{1,}',
            ngram_range=(1, 3), use_idf=1,smooth_idf=1,sublinear_tf=1,
            stop_words = 'english')

# Filling NaNs with empty string
movies_clean['overview'] = movies_clean['overview'].fillna('')

deansublett

scored = movies_clean.sort_values('score', ascending=False)

plt.figure(figsize=(16,6))

ax = sns.barplot(x=scored['score'].head(10), y=scored['original_title'].head(10), data=scored, palette='deep')

#plt.xlim(3.55, 5.25)
plt.title('Best Rated & Most Popular Blend', weight='bold')
plt.xlabel('Score', weight='bold')
plt.ylabel('Movie Title', weight='bold')

deansublett

# My own recommender system
# half/half recommendation based on scaled weighted average & popularity score

from sklearn import preprocessing

min_max_scaler = preprocessing.MinMaxScaler()
movies_scaled = min_max_scaler.fit_transform(movies_clean[['weighted_average', 'popularity']])
movies_norm = pd.DataFrame(movies_scaled, columns=['weighted_average', 'popularity'])
movies_norm.head()

deansublett

popular = movies_ranked.sort_values('popularity', ascending=False)

plt.figure(figsize=(16,6))

ax = sns.barplot(x=popular['popularity'].head(10), y=popular['original_title'].head(10), data=popular, palette='deep')

plt.title('"Most Popular" Movies by TMDB Votes', weight='bold')
plt.xlabel('Popularity Score', weight='bold')
plt.ylabel('Movie Title', weight='bold')

deansublett

import matplotlib.pyplot as plt
import seaborn as sns

wavg = movies_ranked.sort_values('weighted_average', ascending=False)

plt.figure(figsize=(16,6))

ax = sns.barplot(x=wavg['weighted_average'].head(10), y=wavg['original_title'].head(10), data=wavg, palette='deep')

plt.xlim(6.75, 8.35)

deansublett

V = movies_clean['vote_count']
R = movies_clean['vote_average']
C = movies_clean['vote_average'].mean()
m = movies_clean['vote_count'].quantile(0.70)

movies_clean['weighted_average'] = (V/(V+m) * R) + (m/(m+V) * C)

deansublett