diff --git a/recommendation.py b/recommendation.py
new file mode 100644
index 00000000..355064f6
--- /dev/null
+++ b/recommendation.py
@@ -0,0 +1,27 @@
+import pandas as pd
+from sklearn.metrics.pairwise import cosine_similarity
+import numpy as np
+
+def load_data():
+    # Load interaction data
+    data = pd.read_csv('user_interactions.csv')
+    return data
+
+def build_model(data):
+    # Create user-item interaction matrix
+    user_item_matrix = data.pivot_table(index='user_id', columns='item_id', values='interaction', aggfunc='count', fill_value=0)
+    
+    # Compute cosine similarity between users
+    user_similarity = cosine_similarity(user_item_matrix)
+    
+    return user_item_matrix, user_similarity
+
+def recommend_items(user_id, user_item_matrix, user_similarity, num_recommendations=5):
+    user_idx = user_item_matrix.index.get_loc(user_id)
+    similar_users = list(enumerate(user_similarity[user_idx]))
+    similar_users = sorted(similar_users, key=lambda x: x[1], reverse=True)
+    
+    recommendations = []
+    for user, score in similar_users:
+        if user != user_idx:
+            similar_user_items = user_item_matrix.iloc