Update Training Model

WebServer/AIPython/Target_Close_Average.txt

@@ -0,0 +1 @@
+0.198600759951581

WebServer/AIPython/aitrainer.py

@@ -1,8 +1,19 @@
 import pandas as pd
+from sklearn.metrics import f1_score, precision_score, recall_score, accuracy_score
 import features
 import joblib
 import os
+
+# Suppress TensorFlow INFO and WARNING logs
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' 
+
+# Suppress specialized XLA and autotuning logs
+os.environ['TF_XLA_FLAGS'] = '--tf_xla_enable_xla_devices=false'
+os.environ['TF_CPP_MAX_VLOG_LEVEL'] = '0'
+
+# CPU only
 os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
+
 from sklearn.model_selection import train_test_split
 from keras import Sequential, layers, optimizers
 from keras.callbacks import ReduceLROnPlateau
@@ -19,6 +30,9 @@ def TrainAI():
     # Use external featuers to make sure loaded is the same
     dataset = features.MakeFeatures(dataset)
 
+    # Make sure empty columns dont exist
+    dataset = dataset.dropna()
+
     # Create the X, Y vareables
     X, Y, X_Scaler, Y_Scaler = features.Prepare(dataset)
 
@@ -30,7 +44,7 @@ def TrainAI():
     print(dataset.dtypes)
 
     # Split out the test and train
-    train_features, test_features, train_labels, test_labels = train_test_split(X, Y, test_size=0.2)
+    train_features, test_features, train_labels, test_labels = train_test_split(X, Y, test_size=0.2, shuffle=False) # Keep the training and test data in order as its sequential
 
     # Create the DNN
     dnn_model = Sequential([
@@ -45,9 +59,6 @@ def TrainAI():
         layers.Dense(1, activation='linear')            # DNN layer
     ])
 
-    # Allow negative numbers
-    dnn_model.add(layers.LeakyReLU(alpha=0.01))
-
     # Configure the model
     dnn_model.compile(
         optimizer=optimizers.Adam(learning_rate=0.0001, clipvalue=1.0),
@@ -59,7 +70,7 @@ def TrainAI():
     dnn_model.summary()
 
     # Learning rate reducer
-    reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=5, min_lr=0.0000001)
+    reduce_lr = ReduceLROnPlateau(monitor='loss', factor=0.5, patience=5, min_lr=0.0000001)
 
     # Train the model
     Training_Data = dnn_model.fit( 
@@ -68,8 +79,7 @@ def TrainAI():
         batch_size=64,
         epochs=50,                  # Tuned to the point before overfitting
         verbose=1,                  # Show progress
-        validation_split = 0.2,     # Calculate validation results on 20% of the training data.
-        shuffle=True,
+        shuffle=False,              # Time series data
         callbacks=[reduce_lr]       # Reduce the learning_rate every run
     )
 
@@ -78,6 +88,27 @@ def TrainAI():
         test_features, test_labels, verbose=0
     )
 
+    # Perform test on test data split earlier
+    predictions = dnn_model.predict(test_features)
+
+    # Convert to Binary Direction (The "Signal")
+    y_true_binary = (test_labels > 0).astype(int)
+    y_pred_binary = (predictions > 0).astype(int)
+
+    # Calculate the metrics
+    acc = accuracy_score(y_true_binary, y_pred_binary)
+    prec = precision_score(y_true_binary, y_pred_binary)
+    rec = recall_score(y_true_binary, y_pred_binary)
+    f1 = f1_score(y_true_binary, y_pred_binary)
+
+    # Output the meterics; this gets spit out when training from the python file directly. The C# interop does not output these
+    print(f"Test Loss: {test_results[0]:.4f}")
+    print(f"Test MAE: {test_results[1]:.2%}")
+    print(f"Test Accuracy: {acc:.2%}")
+    print(f"Test Precision: {prec:.2%}")
+    print(f"Test Recall: {rec:.2%}")
+    print(f"F1 Score: {f1:.2%}")
+
     # Save the model
     dnn_model.save(os.path.join(DATA_DIR, "model.keras"))