Merge pull request #162 from COS301-SE-2023/feature/retrain_for_new_u…

…sers

Merge Feature/RetrainForNewUsers into Dev

AI/Koja-AI.py

@@ -1,13 +1,20 @@
+import requests
+import schedule as schedule
 from flask import Flask, request, jsonify
 import numpy as np
 import pandas as pd
 import tensorflow as tf
 import tensorflow_recommenders as tfrs
 import json
+import time
+import schedule
+import datetime
+
 import os
 
 app = Flask(__name__)
 
+
 def load_data_and_models():
     f = open('test.json')
     data = json.load(f)
@@ -58,22 +65,63 @@ def load_data_and_models():
 
     return user_model_index, weekday_model_index, time_frame_model_index
 
+
 user_model_index, weekday_model_index, time_frame_model_index = load_data_and_models()
 
+
+def clean_training_data(training_data):
+    data = json.load(training_data)
+    all_events = []
+
+    for block in data:
+        expanded_events = []
+        for event in block['training'] + block['testing']:
+            for time_frame in event['timeFrame']:
+                new_event = event.copy()
+                new_event['startTime'] = time_frame['first'].strip()
+                new_event['endTime'] = time_frame['second'].strip()
+                new_event['category'] = event['category'].strip()
+                new_event['weekday'] = event['weekday'].strip()
+                new_event['userID'] = event['userID'].strip()
+                new_event['timeFrame'] = f"{new_event['startTime']}-{new_event['endTime']}"
+                expanded_events.append(new_event)
+
+        all_events += expanded_events
+
+        return all_events
+
+
+def retrain_for_new_users(training_data):
+    events_data = clean_training_data(training_data)
+
+
+def auto_scheduler(training_data):
+    now = datetime.datetime.now()
+    next_retrain_time = now.replace(hour=23, minute=59, second=0, microsecond=0)
+    if now > next_retrain_time:
+        next_retrain_time += datetime.timedelta(days=1)
+
+    schedule.every().day.at(next_retrain_time.strftime("%H:%M")).do(retrain_for_new_users(training_data))
+
+    while True:
+        schedule.run_pending()
+        time.sleep(1)
+
+
 @app.route('/recommendations', methods=['POST'])
 def recommend_categories():
     if request.is_json:
         data = request.get_json()
         user_id = data['userID']
 
-        maxOutput = 7
+        max_output = 7
         _, titles = user_model_index(np.array([user_id]))
 
         recommendations = []
         for category_id in titles[0][:10]:
             # Convert TensorFlow EagerTensor to numpy array
             category_id_np = category_id.numpy().decode("utf-8")
-            _, weekdays = weekday_model_index(np.array([category_id_np]), k=maxOutput)
+            _, weekdays = weekday_model_index(np.array([category_id_np]), k=max_output)
             _, time_frames = time_frame_model_index(np.array([category_id_np]))
 
             recommendations.append({
@@ -87,6 +135,20 @@ def recommend_categories():
         return "Request was not JSON", 400
 
 
+@app.route('/training-data')
+def get_training_data():
+    api_url = "koja api endpoint"
+    headers = {'Authorization': 'Bearer YOUR_ACCESS_TOKEN'}
+    response = requests.get(api_url, headers=headers)
+
+    if response.status_code == 200:
+        data = response.json()
+        return jsonify(data)
+    else:
+        return jsonify({'error': 'Failed to fetch data'}), 500
+
+
 if __name__ == "__main__":
+    auto_scheduler(get_training_data())
     port = int(os.getenv("PORT", 6000))
     app.run(host='0.0.0.0', port=port, debug=True)

AI/category_model/fingerprint.pb

@@ -1 +1 @@
-�ޫ�ޏ�������Ϥ��
ݻ������
 �ݻ�����(���������
2

+�ޫ�ޏ�������Ϥ��
ݻ������
 �ݻ�����(ߓ�������
2

AI/main.py

@@ -1,4 +1,5 @@
 import json
+from abc import ABC
 from typing import Dict, Text
 import requests
 import pandas as pd
@@ -129,8 +130,9 @@
     candidates=all_candidates
 ))
 
+
 # # Create a TFRS model
-class CategoryRecommender(tfrs.models.Model):
+class CategoryRecommender(tfrs.models.Model, ABC):
     def __init__(self, user_model, category_model, weekday_model, time_frame_model, task):
         super().__init__()
         self.user_model = user_model
@@ -154,6 +156,7 @@ def compute_loss(self, features, training=False):
         # Combine these losses
         return user_loss + category_loss + weekday_loss + time_frame_loss
 
+
 model = CategoryRecommender(user_model, category_model, weekday_model, time_frame_model, task)
 
 # # Configure and train the model
@@ -196,3 +199,4 @@ def compute_loss(self, features, training=False):
 category_model.save("category_model")
 weekday_model.save("weekday_model")
 time_frame_model.save("time_frame_model")
+

AI/time_frame_model/fingerprint.pb

@@ -1 +1 @@
-���������
��ڈ��㬁
Ͻ��IJۋ �׿쩉���
(ז��Ө��O2

+���������
��ڈ��㬁
Ͻ��IJۋ �׿쩉���
(��Ђ���ۅ
2

AI/user_model/fingerprint.pb

@@ -1 +1 @@
-��������qՃ������M���Ɏ����
 �ļ�ˮ���
(�񰏍ܮ��
2

+��������qՃ������M���Ɏ����
 �ļ�ˮ���
(�������r2

AI/weekday_model/fingerprint.pb

@@ -1 +1 @@
-ꓱ�����\�����▭�
���������
 ��������
(������ò=2

+ꓱ�����\�����▭�
���������
 ��������
(‐��Я�D2