updated

app.py

@@ -54,3 +54,36 @@ if st.button("Get Recommendations"):
     else:
         st.write("No sufficient data to generate recommendations.")
 
+
+
+
+import pandas as pd
+from surprise import SVD, Dataset, Reader
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.metrics.pairwise import cosine_similarity
+
+# Load the dataset
+df_songsDB = pd.read_csv('song_dataset.csv')
+
+# Prepare surprise data
+reader = Reader(rating_scale=(1, df_songsDB['play_count'].max()))
+surpriseData = Dataset.load_from_df(df_songsDB[['user', 'song', 'play_count']], reader)
+trainset = surpriseData.build_full_trainset()
+
+# Train SVD model
+algo_SVD = SVD()
+algo_SVD.fit(trainset)
+
+# Define recommendation functions
+def recommend_songs(user_history):
+    user_songs = df_songsDB[df_songsDB['user'].isin(user_history)]['song']
+    unlistened_songs = df_songsDB[~df_songsDB['song'].isin(user_songs)]
+    
+    # Content-based similarity
+    tfidf = TfidfVectorizer()
+    tfidf_matrix = tfidf.fit_transform(unlistened_songs['title'] + " " + unlistened_songs['artist_name'])
+    user_vector = tfidf.transform(user_songs['title'] + " " + user_songs['artist_name'])
+    
+    cosine_sim = cosine_similarity(user_vector, tfidf_matrix)
+    recommendations = unlistened_songs.iloc[cosine_sim.argmax(axis=1)]
+    return recommendations

app_v2.py

@@ -0,0 +1,56 @@
+import streamlit as st
+import pandas as pd
+from sklearn.metrics.pairwise import cosine_similarity
+from sklearn.feature_extraction.text import TfidfVectorizer
+import pickle
+import numpy as np
+import random
+
+# Load the dataset
+df_songsDB = pd.read_csv('song_dataset.csv')
+
+# Build a dropdown for user selection
+st.title("Song Recommendation System")
+user_id = st.selectbox("Select User", options=df_songsDB['user'].unique())
+
+# Dynamically update the dropdown based on user selection
+if user_id:
+    listened_songs = df_songsDB[df_songsDB['user'] == user_id]['title'].unique()
+    selected_songs = st.multiselect("Select Songs You've Liked", options=listened_songs)
+
+# Collaborative Filtering Recommendation
+if st.button("Get Recommendations"):
+    # Prepare user-item interaction matrix
+    user_item_matrix = df_songsDB.pivot_table(
+        index='user', columns='title', values='play_count', fill_value=0
+    )
+
+    # Compute cosine similarity
+    if user_id in user_item_matrix.index:
+        user_vector = user_item_matrix.loc[user_id].values.reshape(1, -1)
+        similarity = cosine_similarity(user_vector, user_item_matrix.values)
+        recommendations = (
+            user_item_matrix.columns[np.argsort(-similarity[0])].tolist()
+        )
+
+        # Filter out already listened songs
+        recommendations = [
+            song for song in recommendations if song not in listened_songs
+        ]
+        
+        # Shuffle recommendations to mix up starting letters
+        random.shuffle(recommendations)
+        
+        # Select 10 random recommendations if there are enough songs
+#        if len(recommendations) >= 10:
+#            random_recommendations = random.sample(recommendations, 10)
+#        else:
+        random_recommendations = recommendations
+
+        # Display Recommendations
+        st.write("Recommended Songs for You:")
+        for song in random_recommendations:
+            st.write(song)
+    else:
+        st.write("No sufficient data to generate recommendations.")
+

requirements.txt

@@ -2,3 +2,5 @@ streamlit>=1.4.0
 pandas>=1.0.0
 scikit-learn>=1.0.0
 numpy>=1.19.5
+
+surprise

requirements_v2.txt

@@ -0,0 +1,4 @@
+streamlit>=1.4.0
+pandas>=1.0.0
+scikit-learn>=1.0.0
+numpy>=1.19.5