Merge pull request #259 from COS301-SE-2024/feature/backend/Model_Dep…

…loyment

Feature/backend/model deployment

datasets/processor.py

@@ -0,0 +1,32 @@
+import pandas as pd
+
+# Define the levels based on capacity
+def get_level(capacity):
+    if capacity >= 1600:
+        return 7
+    elif capacity >= 1400:
+        return 6
+    elif capacity >= 1200:
+        return 5
+    elif capacity >= 1000:
+        return 4
+    elif capacity >= 800:
+        return 3
+    elif capacity >= 600:
+        return 2
+    else:
+        return 1
+
+# Read the CSV data into a DataFrame
+file_path = 'datasets/Attendance_data(1).csv'
+data = pd.read_csv(file_path)
+
+df = pd.DataFrame(data)
+
+# Apply the get_level function to create the Level column
+df['Level'] = df['Number_Attended'].apply(get_level)
+
+# Save the modified DataFrame back to a CSV file
+df.to_csv('output.csv', index=False)
+
+print(df)

maestro/LSTM(1).py

@@ -216,5 +216,9 @@ def predict_weekly_attendance(month, start_day):
 plt.show()
 
 # Save the model in the SavedModel format
-model.save('attendance_model')
+tf.saved_model.save(model, 'attendance_model/1')
 # model.export('C:/Users/retha/Capstone/occupi/models/attendance_model/1')
+
+new_model = tf.keras.models.load_model('C:/Users/retha/Capstone/occupi/attendance_model.keras')
+new_model.summary()
+tf.saved_model.save(new_model, 'serving/')  # Save the model in the SavedModel format

maestro/test.py

@@ -4,7 +4,7 @@
 from sklearn.preprocessing import StandardScaler
 
 # Define the URL for the TensorFlow Serving API
-url = 'http://localhost:8503/v1/models/attendance_model:predict'
+url = 'http://localhost:8501/v1/models/attendance_model:predict'
 
 # Define the attendance levels based on the bin ranges
 attendance_levels = ["0-300", "300-600", "600-900", "900-1200", "1200-1500", "1500-1800", "1800+"]

python-code/app.py

@@ -1,23 +1,25 @@
-# app.py
 from flask import Flask, request, jsonify
 import logging
-from prediction import get_prediction
-from scaler import scaler
 from datetime import datetime, timedelta
+from prediction import get_prediction
+import joblib
 
 # Initialize the Flask application
 app = Flask(__name__)
 
+# Load the scaler
+scaler = joblib.load('attendance_scaler.pkl')
+
 # Function to determine if a given date is a weekend
 def is_weekend(date):
     return date.weekday() >= 5
 
 # Function to determine if a given date is a special event (placeholder logic)
 def is_special_event(date):
-    special_events_dates = [(7, 4), (12, 25),(4,23),(12,9),(8,5),(3,6),(11,27),(3,10),(7,26)]  # Example: 4th of July, Christmas
+    special_events_dates = [(7, 4), (12, 25), (4, 23), (12, 9), (8, 5), (3, 6), (11, 27), (3, 10), (7, 26)]  # Example: 4th of July, Christmas
     return 1 if (date.month, date.day) in special_events_dates else 0
 
-@app.route('/predict', methods=['POST'])
+@app.route('/predict', methods=['GET'])
 def predict():
     try:
         # Get current date
@@ -29,8 +31,8 @@ def predict():
         day_of_month = current_date.day
         weekend = is_weekend(current_date)
         special_event = is_special_event(current_date)
-         # Set factor based on special event
-        factor = 1.5 if special_event else 1.0
+        # Set factor based on special event
+        # factor = 1.5 if special_event else 1.0
 
         predicted_class, predicted_attendance_level = get_prediction(day_of_week, month, day_of_month, weekend, special_event, scaler, factor)
 
@@ -47,7 +49,7 @@ def predict():
         logging.error(f"Error in predict endpoint: {str(e)}")
         return jsonify({"error": str(e)}), 500
 
-@app.route('/predict_week', methods=['POST'])
+@app.route('/predict_week', methods=['GET'])
 def predict_week():
     try:
         # Get the current date
@@ -68,8 +70,8 @@ def predict_week():
             weekend = is_weekend(date)
             special_event = is_special_event(date)
 
-           # Set factor based on special event
-            factor = 1.5 if special_event else 1.0
+            # Set factor based on special event
+            # factor = 1.5 if special_event else 1.0
 
             # Get prediction
             predicted_class, predicted_attendance_level = get_prediction(day_of_week, month, day_of_month, weekend, special_event, scaler, factor)
@@ -92,4 +94,4 @@ def predict_week():
         return jsonify({"error": str(e)}), 500
 
 if __name__ == '__main__':
-    app.run(debug=True, host='0.0.0.0')
+    app.run(debug=True, host='0.0.0.0', port=9000)

python-code/docker-compose.yml

@@ -0,0 +1,18 @@
+version: '3.8'
+
+services:
+  web:
+    build: .
+    ports:
+      - "9000:9000"
+    depends_on:
+      - model
+
+  model:
+    image: tensorflow/serving:latest
+    environment:
+      - MODEL_NAME=attendance_model
+    volumes:
+      - ../models/attendance_model:/models/attendance_model
+    ports:
+      - "8501:8501"

python-code/prediction.py

@@ -6,7 +6,7 @@
 import logging
 
 # Define the URL for the TensorFlow Serving API
-url = 'http://localhost:8501/v1/models/attendance_model:predict'
+url = 'http://model:8501/v1/models/attendance_model:predict'
 
 # Define the attendance levels based on the bin ranges
 attendance_levels = ["0-300", "300-600", "600-900", "900-1200", "1200-1500", "1500-1800", "1800+"]