amazon recommendation.py

counts = df['user_id'].value_counts()
data = df[df['user_id'].isin(counts[counts >= 50].index)]
data.groupby('product_id')['ratings'].mean().sort_values(ascending=False) 
final_ratings = data.pivot(index = 'user_id', columns ='product_id', values = 'ratings').fillna(0)

num_of_ratings = np.count_nonzero(final_ratings)
possible_ratings = final_ratings.shape[0] * final_ratings.shape[1]
density = (num_of_ratings/possible_ratings)
density *= 100
final_ratings_T = final_ratings.transpose()

grouped = data.groupby('product_id').agg({'user_id': 'count'}).reset_index()
grouped.rename(columns = {'user_id': 'score'},inplace=True)
training_data = grouped.sort_values(['score', 'product_id'], ascending = [0,1]) 
training_data['Rank'] = training_data['score'].rank(ascending=0, method='first') 
recommendations = training_data.head()