diff --git a/.github/workflows/fetch-data.yml b/.github/workflows/fetch-data.yml
index 7c0f19d2..fc50a490 100644
--- a/.github/workflows/fetch-data.yml
+++ b/.github/workflows/fetch-data.yml
@@ -7,6 +7,8 @@ on:
 
 jobs:
   Fetch-Route-Data:
+    permissions:
+      contents: write
     runs-on: ubuntu-latest
 
     steps:
@@ -15,7 +17,7 @@ jobs:
       - name: Setup Python environment
         uses: actions/setup-python@v4
         with:
-            python-version: '3.8'
+            python-version: '3.12'
             architecture: 'x64'
             cache: 'pip'
             cache-dependency-path: crawling/requirements.txt
@@ -45,6 +47,7 @@ jobs:
           python ./crawling/matchGtfs.py
           python ./crawling/cleansing.py
           python ./crawling/mergeRoutes.py
+          python ./crawling/groupBus.py
           python ./crawling/routeCompare.py
           python ./crawling/mtrExits.py
 
@@ -67,12 +70,21 @@ jobs:
             ROUTE_BUS.xml
             route-ts/
             exits.mtr.json
+            groupBus_all.json
       - name: Update MD5
         run: md5sum routeFareList.min.json | cut -f1 -d ' ' | tr -d $'\n' > routeFareList.md5
       - name: create deployment folder
         run: mkdir -p build
       - name: cp files into deployment folder
-        run: cp -r routeFareList.json routeFareList.min.json routeFareList.md5 CNAME exits.mtr.json route-ts build/
+        run: |
+          cp -r \
+          routeFareList.json \
+          routeFareList.min.json \
+          routeFareList.md5 \
+          CNAME \
+          exits.mtr.json \
+          groupBus_all.json \
+          build/
       - name: cp route-ts into deployment folder
         run: cp -r route-ts build
       - name: Update resources
diff --git a/crawling/groupBus.py b/crawling/groupBus.py
new file mode 100644
index 00000000..53322a34
--- /dev/null
+++ b/crawling/groupBus.py
@@ -0,0 +1,92 @@
+from scipy.spatial import KDTree
+import json
+import math
+import polars as pl
+
+def haversine_distance(lat1, lon1, lat2, lon2):
+    R = 6371000  # Earth radius in meters
+    
+    phi1 = math.radians(lat1)
+    phi2 = math.radians(lat2)
+    delta_phi = math.radians(lat2 - lat1)
+    delta_lambda = math.radians(lon2 - lon1)
+    
+    a = math.sin(delta_phi/2)**2 + math.cos(phi1) * math.cos(phi2) * math.sin(delta_lambda/2)**2
+    c = 2 * math.atan2(math.sqrt(a), math.sqrt(1-a))
+    
+    return R * c
+
+def calculate_bearing(lat1, lon1, lat2, lon2):
+    phi1 = math.radians(lat1)
+    phi2 = math.radians(lat2)
+    delta_lambda = math.radians(lon2 - lon1)
+    
+    y = math.sin(delta_lambda) * math.cos(phi2)
+    x = math.cos(phi1) * math.sin(phi2) - math.sin(phi1) * math.cos(phi2) * math.cos(delta_lambda)
+    theta = math.atan2(y, x)
+    
+    return (math.degrees(theta) + 360) % 360
+
+def group_bus_stops(bus_stops, max_distance=50, bearing_threshold=35):
+    tree = KDTree(bus_stops.select("lat", "lng"))
+    groups = pl.DataFrame(schema={"id":str, "lat":pl.Float64, "lng":pl.Float64, "name_en":str, "name_zh":str, "bus_group_id":pl.Int32})
+    visited = set()
+    group_id = 1
+
+    print(len(bus_stops))
+    
+    for i in range(len(bus_stops)):
+        if i in visited:
+            continue
+
+        # Create a new group for this stop
+        stop1 = bus_stops[i]
+        stop1 = stop1.with_columns(
+            bus_group_id=group_id
+        )
+        group = stop1
+        nearby_stop_indices = tree.query_ball_point([stop1['lat'][0], stop1['lng'][0]], r=max_distance/1000)
+        
+        for j in nearby_stop_indices:
+            if i != j and j not in visited:
+                stop2 = bus_stops[j]
+
+                distance = haversine_distance(stop1['lat'][0], stop1['lng'][0], stop2['lat'][0], stop2['lng'][0])
+                
+                if distance <= max_distance:
+                    if group.height > 1:
+                        prev_stop = group[-2]
+                        bearing1 = calculate_bearing(prev_stop['lat'][0], prev_stop['lng'][0], stop1['lat'][0], stop1['lng'][0])
+                        bearing2 = calculate_bearing(stop1['lat'][0], stop1['lng'][0], stop2['lat'][0], stop2['lng'][0])
+                        
+                        if abs(bearing1 - bearing2) <= bearing_threshold or abs(bearing1 - bearing2) >= 360 - bearing_threshold:
+                            stop2 = stop2.with_columns(
+                                bus_group_id=group_id
+                            )
+                            group = group.vstack(stop2)
+                    else:
+                        stop2 = stop2.with_columns(
+                            bus_group_id=group_id
+                        )
+                        group = group.vstack(stop2)
+        
+        group_id += 1
+        visited.add(i)
+        groups = groups.vstack(group)
+    
+    return groups
+
+if __name__ == '__main__': 
+    with open("routeFareList.min.json", 'r', encoding='utf8') as f:
+        r = json.load(f)
+        r = r['stopList']
+
+    j2 = [{"id": id, "lat": v['location']['lat'], "lng": v['location']['lng'],
+           "name_en": v['name']['en'], "name_zh": v['name']['zh']} for id, v in r.items()]
+
+    df = pl.from_dicts(j2) #.lazy()
+    #df = df.filter(pl.col('name_zh').str.contains('宋皇'))
+    grouped_bus_stops = group_bus_stops(df)
+
+    with open(f'groupBus_all.json', 'w', encoding='utf8') as f:
+         f.write(grouped_bus_stops.write_json())
\ No newline at end of file
diff --git a/crawling/parseGtfs.py b/crawling/parseGtfs.py
index b1a95384..f51ef312 100644
--- a/crawling/parseGtfs.py
+++ b/crawling/parseGtfs.py
@@ -97,7 +97,7 @@ async def parseGtfs():
       _tmp.sort(key=takeFirst)
       routeList[route_id]['fares'][bound] = [v[0] for k,v in _tmp]
 
-  nameReg = re.compile('\[(.*)\] (.*)')
+  nameReg = re.compile(r'\[(.*)\] (.*)')
   def parseStopName(name):
     ret = {}
     for str in name.split('|'):
diff --git a/crawling/parseGtfsEn.py b/crawling/parseGtfsEn.py
index 42436557..c67d5ce3 100644
--- a/crawling/parseGtfsEn.py
+++ b/crawling/parseGtfsEn.py
@@ -97,7 +97,7 @@ async def parseGtfs():
       _tmp.sort(key=takeFirst)
       routeList[route_id]['fares'][bound] = [v[0] for k,v in _tmp]
 
-  nameReg = re.compile('\[(.*)\] (.*)')
+  nameReg = re.compile(r'\[(.*)\] (.*)')
   def parseStopName(name):
     ret = {}
     for str in name.split('|'):
diff --git a/crawling/requirements.txt b/crawling/requirements.txt
index a76cdd79..49d263c2 100644
--- a/crawling/requirements.txt
+++ b/crawling/requirements.txt
@@ -12,7 +12,9 @@ PySocks==1.7.1
 six==1.15.0
 urllib3==1.26.4
 wheel==0.36.2
-pyproj==3.3.0
+pyproj==3.6.1
 httpx==0.25.2
 xxhash==3.2.0
+polars==1.7.1
+scipy==1.14.1
 -e .