Update Interface_Graphique/interface_graphique/services/model_service.py

Interface_Graphique/interface_graphique/services/model_service.py

@@ -1,12 +1,54 @@
 import numpy as np
+import joblib
+from tensorflow.keras.models import load_model
 
+# =========================
+# LOAD MODEL + PREPROCESS
+# =========================
+model = load_model("models/global_return_lstm.keras")
+scaler = joblib.load("models/return_scaler.save")
+encoder = joblib.load("models/symbol_encoder.save")
+
+SEQ_LEN = 60
+SIGNAL_THRESHOLD = 0.001
+
+
+# =========================
+# PREPARE INPUT
+# =========================
+def prepare_input(symbol, returns):
+
+    # encode symbol
+    symbol_id = encoder.transform([symbol])[0]
+
+    # scale returns
+    returns = returns.reshape(-1, 1)
+    returns_scaled = scaler.transform(returns)
+
+    # reshape for LSTM
+    X_price = returns_scaled.reshape(1, SEQ_LEN, 1)
+    X_symbol = np.array([[symbol_id]])
+
+    return X_price, X_symbol
+
+
+# =========================
+# PREDICT SIGNAL
+# =========================
 def predict_signal(symbol, returns):
 
-    avg = np.mean(returns)
+    if len(returns) < SEQ_LEN:
+        return "HOLD"
+
+    returns = returns[-SEQ_LEN:]
+
+    X_price, X_symbol = prepare_input(symbol, returns)
+
+    pred_return = model.predict([X_price, X_symbol], verbose=0)[0][0]
 
-    if avg > 0:
+    if pred_return > SIGNAL_THRESHOLD:
         return "BUY"
-    elif avg < 0:
+    elif pred_return < -SIGNAL_THRESHOLD:
         return "SELL"
     else:
         return "HOLD"