adds new recommender and improves old ones

pages/{1 - Popularity based recommender.py → 1 - Popularity-Based Recommender.py}

@@ -10,7 +10,7 @@ st.title("Popularity-Based Recommender")
 st.write("""
 ### Instructions
 Move the slider to the desired number of recommendations you wish to receive. 
-Afterwards, simply click the "Recommend!" button to receive recommendations of the most popular movies in our database. 
+Afterwards, simply click the "Get Recommendations" button to receive recommendations of the most popular movies in our database. 
 If you want, you can narrow it down by picking one or several genre(s).  
 """)
 
@@ -43,5 +43,5 @@ genres = st.multiselect('Optional: Select one or more genres', genre_list, defau
 genres_regex = transform_genre_to_regex(genres)
 
 # EXECUTION:
-if st.button("Recommend!"):
+if st.button("Get Recommendations"):
     st.write(get_popular_recommendations(number_of_recommendations, genres_regex))

pages/{2 - User based recommender.py → 2 - User-Based Recommender.py}

@@ -23,7 +23,8 @@ st.title("User-Based Recommender")
 st.write("""
 ### Instructions
 Type in the user-ID you want to receive recommendations for.  
-Move the slider to the desired number of recommendations you wish to receive. 
+Move the slider to the desired number of recommendations you wish to receive.  
+Afterwards, simply click the "Get Recommendations" button to receive recommendations that are most suitable for the given user.  
 """)
 st.write("""
 Optional: You can narrow down the recommendations by picking one or several genre(s).  
@@ -51,9 +52,9 @@ def get_user_recommendations(user_id, n, genres):
     recommendations = (
         weighted_averages
             .sort_values("predicted_rating", ascending=False)
-            .head(n)
             .merge(movies, how= 'left', left_index = True, right_on = 'title')
             [lambda df: df["genres"].str.contains(genres, regex=True)]
+            .head(n)
     )
 
     return recommendations[['title', 'genres']]
@@ -74,5 +75,5 @@ genres = st.multiselect('Optional: Select one or more genres', genre_list, defau
 genres_regex = transform_genre_to_regex(genres)
 
 # EXECUTION:
-if st.button("Recommend!"):
+if st.button("Get Recommendations"):
     st.write(get_user_recommendations(user_id_input, number_of_recommendations, genres_regex))

pages/3 - Similarity-Based Recommender.py

@@ -0,0 +1,84 @@
+import streamlit as st
+import pandas as pd
+from sklearn.metrics.pairwise import cosine_similarity
+import re
+
+# DATA:
+movies = pd.read_csv('https://raw.githubusercontent.com/tobiasaurer/recommender-systems/main/movie_data/movies.csv')
+ratings = pd.read_csv('https://raw.githubusercontent.com/tobiasaurer/recommender-systems/main/movie_data/ratings.csv')
+
+# create "database" to use for recommendations
+movie_user_matrix = (
+                ratings
+                    .merge(movies, on='movieId')[['title', 'rating', 'userId']]
+                    .pivot_table(index='title', columns='userId', values='rating')
+                    .fillna(0)
+                )
+similarities_movies = pd.DataFrame(cosine_similarity(movie_user_matrix),
+                                  index=movie_user_matrix.index,
+                                  columns=movie_user_matrix.index)
+
+# INSTRUCTIONS:
+st.title("User-Based Recommender")
+st.write("""
+### Instructions
+Type in the title of a movie for which you would like to receive similar recommendations.  
+Move the slider to the desired number of recommendations you wish to receive.  
+Afterwards, simply click the "Get Recommendations" button to receive recommendations that are most similar to the given movie. 
+""")
+st.write("""
+Optional: You can narrow down the recommendations by picking one or several genre(s).  
+  
+Note: The more genres you choose, the fewer movies will be recommended.
+""")
+
+# FUNCTIONS:
+
+def get_similar_recommendations(movie_title, n, genres):
+    
+    # select similarity for chosen movie
+    similarities = pd.DataFrame(similarities_movies.loc[similarities_movies.index != movie_title, movie_title])
+ 
+    # exclude genres if necessary and return the n movies with the highest similarity
+    recommendations = (
+        similarities
+            .merge(movies, how= 'left', left_index = True, right_on = 'title')
+            [lambda df: df["genres"].str.contains(genres, regex=True)]
+            .head(n)
+            [['title', 'genres']]
+            )
+
+    return recommendations
+
+def transform_genre_to_regex(genres):
+    regex = ""
+    for genre in genres:
+        regex += f"(?=.*{genre})"
+    return regex
+
+def find_movie_title(user_input):
+    title_list = movies.title.unique()
+    
+    r = re.compile(f".*{user_input}.*")
+    result = []
+
+    for title in title_list:
+        match = r.findall(title)
+        if match:
+            result.append(match)
+    
+    return result[0][0]
+
+# USER INPUT:
+movie_title_raw = st.text_input('Movie Title')
+movie_title = find_movie_title(movie_title_raw)
+
+number_of_recommendations = st.slider("Number of recommendations", 1, 10, 5)
+
+genre_list = set([inner for outer in movies.genres.str.split('|') for inner in outer])
+genres = st.multiselect('Optional: Select one or more genres', genre_list, default=None, key=None, help=None, on_change=None, args=None, kwargs=None, disabled=False)
+genres_regex = transform_genre_to_regex(genres)
+
+# EXECUTION:
+if st.button("Get Recommendations"):
+    st.write(get_similar_recommendations(movie_title, number_of_recommendations, genres_regex))

requirements.txt

@@ -1,3 +1,4 @@
 streamlit
 pandas
-scikit-learn
+scikit-learn
+re