Spaces:
Paused
Paused
Update ml_engine/oracle_engine.py
Browse files- ml_engine/oracle_engine.py +64 -39
ml_engine/oracle_engine.py
CHANGED
|
@@ -16,10 +16,7 @@ CONFIDENCE_THRESHOLD = 0.65
|
|
| 16 |
class OracleEngine:
|
| 17 |
def __init__(self, model_dir: str = "ml_models/Unified_Models_V1"):
|
| 18 |
"""
|
| 19 |
-
Oracle V4.
|
| 20 |
-
- يحلل الاتجاه (صعود/هبوط).
|
| 21 |
-
- يمرر فقط فرص الصعود (Long) للتنفيذ.
|
| 22 |
-
- يحجب فرص الهبوط (Short) ويعتبرها "مخاطرة" (WAIT).
|
| 23 |
"""
|
| 24 |
self.model_dir = model_dir
|
| 25 |
self.model_direction = None
|
|
@@ -28,7 +25,7 @@ class OracleEngine:
|
|
| 28 |
|
| 29 |
self.feature_cols = []
|
| 30 |
self.initialized = False
|
| 31 |
-
print("🧠 [Oracle V4
|
| 32 |
|
| 33 |
async def initialize(self):
|
| 34 |
"""تحميل النماذج وخريطة الميزات"""
|
|
@@ -65,24 +62,30 @@ class OracleEngine:
|
|
| 65 |
print(f"❌ [Oracle] Init Error: {e}")
|
| 66 |
return False
|
| 67 |
|
| 68 |
-
# ==========================================================================
|
| 69 |
-
# 🛠️ هندسة الميزات (مطابقة لـ DataFactory)
|
| 70 |
-
# ==========================================================================
|
| 71 |
def _calculate_snapshot_features(self, df, tf_prefix):
|
| 72 |
df = df.copy()
|
| 73 |
df['close'] = df['close'].astype(float)
|
| 74 |
df['volume'] = df['volume'].astype(float)
|
| 75 |
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
def _create_feature_vector(self, ohlcv_data: Dict[str, Any], titan_score: float, mc_score: float, pattern_score: float) -> Optional[pd.DataFrame]:
|
| 88 |
try:
|
|
@@ -106,6 +109,8 @@ class OracleEngine:
|
|
| 106 |
feats_4h = pd.DataFrame(np.zeros((1, 4)), columns=[f'4h_{c}' for c in ['slope', 'rsi', 'atr_pct', 'vol_z']])
|
| 107 |
|
| 108 |
vector = pd.concat([feats_1h, feats_15m, feats_4h], axis=1)
|
|
|
|
|
|
|
| 109 |
vector['sim_titan_score'] = float(titan_score)
|
| 110 |
vector['sim_mc_score'] = float(mc_score)
|
| 111 |
vector['sim_pattern_score'] = float(pattern_score)
|
|
@@ -113,13 +118,16 @@ class OracleEngine:
|
|
| 113 |
final_vector = pd.DataFrame(columns=self.feature_cols)
|
| 114 |
for col in self.feature_cols:
|
| 115 |
if col in vector.columns:
|
| 116 |
-
|
|
|
|
|
|
|
| 117 |
else:
|
| 118 |
final_vector.at[0, col] = 0.0
|
| 119 |
|
| 120 |
return final_vector.astype(float)
|
| 121 |
|
| 122 |
except Exception as e:
|
|
|
|
| 123 |
return None
|
| 124 |
|
| 125 |
# ==========================================================================
|
|
@@ -128,38 +136,50 @@ class OracleEngine:
|
|
| 128 |
async def predict(self, symbol_data: Dict[str, Any]) -> Dict[str, Any]:
|
| 129 |
"""تحليل الفرصة: هل هي صالحة للشراء (SPOT)؟"""
|
| 130 |
if not self.initialized:
|
| 131 |
-
return {'action': 'WAIT', 'reason': 'Not initialized'}
|
| 132 |
|
| 133 |
try:
|
| 134 |
ohlcv = symbol_data.get('ohlcv')
|
| 135 |
current_price = symbol_data.get('current_price', 0.0)
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 139 |
|
| 140 |
features = self._create_feature_vector(ohlcv, titan, mc, patt)
|
| 141 |
if features is None:
|
| 142 |
-
return {'action': 'WAIT', 'reason': 'Features failed'}
|
| 143 |
|
| 144 |
# 1. التنبؤ بالاتجاه (Direction)
|
| 145 |
# 0=Long (Buy), 1=Short (Drop/Avoid)
|
| 146 |
dir_probs = self.model_direction.predict(features)[0]
|
| 147 |
|
| 148 |
if isinstance(dir_probs, (np.ndarray, list)):
|
| 149 |
-
prob_long = dir_probs[0]
|
| 150 |
-
prob_short = dir_probs[1]
|
| 151 |
else:
|
| 152 |
-
|
| 153 |
-
|
|
|
|
|
|
|
|
|
|
| 154 |
|
| 155 |
# --- [SPOT LOGIC ENFORCEMENT] ---
|
| 156 |
-
#
|
| 157 |
-
#
|
|
|
|
| 158 |
if prob_short > prob_long:
|
| 159 |
return {
|
| 160 |
'action': 'WAIT',
|
| 161 |
-
'reason': f'Bearish
|
| 162 |
-
'direction': 'SHORT'
|
|
|
|
|
|
|
| 163 |
}
|
| 164 |
|
| 165 |
# إذا وصلنا هنا، فالاتجاه هو LONG (شراء)
|
|
@@ -169,8 +189,10 @@ class OracleEngine:
|
|
| 169 |
if confidence < CONFIDENCE_THRESHOLD:
|
| 170 |
return {
|
| 171 |
'action': 'WAIT',
|
| 172 |
-
'reason': f'Low
|
| 173 |
-
'direction': 'LONG'
|
|
|
|
|
|
|
| 174 |
}
|
| 175 |
|
| 176 |
# 3. التنبؤ بالأهداف والقوة
|
|
@@ -187,8 +209,11 @@ class OracleEngine:
|
|
| 187 |
tp_labels = ['TP1', 'TP2', 'TP3', 'TP4']
|
| 188 |
target_profile = tp_labels[tp_class_idx]
|
| 189 |
|
| 190 |
-
# 4. حساب المستويات السعرية
|
| 191 |
atr_pct_val = features['1h_atr_pct'].iloc[0]
|
|
|
|
|
|
|
|
|
|
| 192 |
atr_abs = atr_pct_val * current_price
|
| 193 |
|
| 194 |
tp_map = {
|
|
@@ -199,11 +224,11 @@ class OracleEngine:
|
|
| 199 |
}
|
| 200 |
|
| 201 |
primary_tp = tp_map[target_profile]
|
| 202 |
-
sl_price = current_price - (1.2 * atr_abs)
|
| 203 |
|
| 204 |
return {
|
| 205 |
-
'action': 'WATCH',
|
| 206 |
-
'action_type': 'BUY',
|
| 207 |
'confidence': float(confidence),
|
| 208 |
'strength': float(strength),
|
| 209 |
'target_class': target_profile,
|
|
@@ -215,4 +240,4 @@ class OracleEngine:
|
|
| 215 |
|
| 216 |
except Exception as e:
|
| 217 |
print(f"❌ [Oracle] Prediction Error: {e}")
|
| 218 |
-
return {'action': 'WAIT', 'reason': 'Error'}
|
|
|
|
| 16 |
class OracleEngine:
|
| 17 |
def __init__(self, model_dir: str = "ml_models/Unified_Models_V1"):
|
| 18 |
"""
|
| 19 |
+
Oracle V4.1: Spot-Only Strategic Brain (Fixed Logging)
|
|
|
|
|
|
|
|
|
|
| 20 |
"""
|
| 21 |
self.model_dir = model_dir
|
| 22 |
self.model_direction = None
|
|
|
|
| 25 |
|
| 26 |
self.feature_cols = []
|
| 27 |
self.initialized = False
|
| 28 |
+
print("🧠 [Oracle V4.1] Engine Instance Created.")
|
| 29 |
|
| 30 |
async def initialize(self):
|
| 31 |
"""تحميل النماذج وخريطة الميزات"""
|
|
|
|
| 62 |
print(f"❌ [Oracle] Init Error: {e}")
|
| 63 |
return False
|
| 64 |
|
|
|
|
|
|
|
|
|
|
| 65 |
def _calculate_snapshot_features(self, df, tf_prefix):
|
| 66 |
df = df.copy()
|
| 67 |
df['close'] = df['close'].astype(float)
|
| 68 |
df['volume'] = df['volume'].astype(float)
|
| 69 |
|
| 70 |
+
# تجنب الأخطاء إذا كانت البيانات قصيرة جداً
|
| 71 |
+
if len(df) < 15:
|
| 72 |
+
return pd.DataFrame()
|
| 73 |
+
|
| 74 |
+
try:
|
| 75 |
+
df[f'{tf_prefix}_slope'] = ta.slope(df['close'], length=7)
|
| 76 |
+
df[f'{tf_prefix}_rsi'] = ta.rsi(df['close'], length=14)
|
| 77 |
+
atr = ta.atr(df['high'], df['low'], df['close'], length=14)
|
| 78 |
+
df[f'{tf_prefix}_atr_pct'] = atr / df['close']
|
| 79 |
+
vol_mean = df['volume'].rolling(20).mean()
|
| 80 |
+
vol_std = df['volume'].rolling(20).std()
|
| 81 |
+
df[f'{tf_prefix}_vol_z'] = (df['volume'] - vol_mean) / (vol_std + 1e-9)
|
| 82 |
+
|
| 83 |
+
cols = [f'{tf_prefix}_slope', f'{tf_prefix}_rsi', f'{tf_prefix}_atr_pct', f'{tf_prefix}_vol_z']
|
| 84 |
+
# استخدام fillna(0) لضمان عدم وجود قيم فارغة
|
| 85 |
+
return df[cols].ffill().bfill().fillna(0)
|
| 86 |
+
except Exception as e:
|
| 87 |
+
# print(f"⚠️ Feature Calc Error ({tf_prefix}): {e}")
|
| 88 |
+
return pd.DataFrame()
|
| 89 |
|
| 90 |
def _create_feature_vector(self, ohlcv_data: Dict[str, Any], titan_score: float, mc_score: float, pattern_score: float) -> Optional[pd.DataFrame]:
|
| 91 |
try:
|
|
|
|
| 109 |
feats_4h = pd.DataFrame(np.zeros((1, 4)), columns=[f'4h_{c}' for c in ['slope', 'rsi', 'atr_pct', 'vol_z']])
|
| 110 |
|
| 111 |
vector = pd.concat([feats_1h, feats_15m, feats_4h], axis=1)
|
| 112 |
+
|
| 113 |
+
# التأكد من أن البيانات الرقمية (Scores) تُمرر بشكل صحيح
|
| 114 |
vector['sim_titan_score'] = float(titan_score)
|
| 115 |
vector['sim_mc_score'] = float(mc_score)
|
| 116 |
vector['sim_pattern_score'] = float(pattern_score)
|
|
|
|
| 118 |
final_vector = pd.DataFrame(columns=self.feature_cols)
|
| 119 |
for col in self.feature_cols:
|
| 120 |
if col in vector.columns:
|
| 121 |
+
val = vector[col].iloc[0]
|
| 122 |
+
# حماية إضافية ضد NaN
|
| 123 |
+
final_vector.at[0, col] = float(val) if not pd.isna(val) else 0.0
|
| 124 |
else:
|
| 125 |
final_vector.at[0, col] = 0.0
|
| 126 |
|
| 127 |
return final_vector.astype(float)
|
| 128 |
|
| 129 |
except Exception as e:
|
| 130 |
+
print(f"❌ Vector Creation Error: {e}")
|
| 131 |
return None
|
| 132 |
|
| 133 |
# ==========================================================================
|
|
|
|
| 136 |
async def predict(self, symbol_data: Dict[str, Any]) -> Dict[str, Any]:
|
| 137 |
"""تحليل الفرصة: هل هي صالحة للشراء (SPOT)؟"""
|
| 138 |
if not self.initialized:
|
| 139 |
+
return {'action': 'WAIT', 'reason': 'Not initialized', 'confidence': 0.0}
|
| 140 |
|
| 141 |
try:
|
| 142 |
ohlcv = symbol_data.get('ohlcv')
|
| 143 |
current_price = symbol_data.get('current_price', 0.0)
|
| 144 |
+
|
| 145 |
+
# استخراج القيم والتأكد منها (DEBUG)
|
| 146 |
+
titan = symbol_data.get('titan_score', 0.0)
|
| 147 |
+
mc = symbol_data.get('mc_score', 0.0)
|
| 148 |
+
patt = symbol_data.get('patterns_score', 0.0)
|
| 149 |
+
|
| 150 |
+
# طباعة تصحيحية لمرة واحدة إذا كانت القيم صفرية بشكل مريب
|
| 151 |
+
if titan == 0 and patt == 0:
|
| 152 |
+
print(f"⚠️ [Oracle Warning] Input scores are ZERO for {symbol_data.get('symbol')}. Check upstream.")
|
| 153 |
|
| 154 |
features = self._create_feature_vector(ohlcv, titan, mc, patt)
|
| 155 |
if features is None:
|
| 156 |
+
return {'action': 'WAIT', 'reason': 'Features failed', 'confidence': 0.0}
|
| 157 |
|
| 158 |
# 1. التنبؤ بالاتجاه (Direction)
|
| 159 |
# 0=Long (Buy), 1=Short (Drop/Avoid)
|
| 160 |
dir_probs = self.model_direction.predict(features)[0]
|
| 161 |
|
| 162 |
if isinstance(dir_probs, (np.ndarray, list)):
|
| 163 |
+
prob_long = float(dir_probs[0])
|
| 164 |
+
prob_short = float(dir_probs[1])
|
| 165 |
else:
|
| 166 |
+
# إذا كانت النتيجة قيمة واحدة (Binary output for Class 1)
|
| 167 |
+
# نفترض أن Class 1 هو Short (حسب تدريبك السابق) أو العكس.
|
| 168 |
+
# حسب الكود الأصلي: prob_short = dir_probs
|
| 169 |
+
prob_short = float(dir_probs)
|
| 170 |
+
prob_long = 1.0 - prob_short
|
| 171 |
|
| 172 |
# --- [SPOT LOGIC ENFORCEMENT] ---
|
| 173 |
+
# التصحيح: نعيد قيمة confidence حتى لو كانت الحالة WAIT
|
| 174 |
+
# لكي تظهر في السجلات (Logs)
|
| 175 |
+
|
| 176 |
if prob_short > prob_long:
|
| 177 |
return {
|
| 178 |
'action': 'WAIT',
|
| 179 |
+
'reason': f'Bearish (Short Prob: {prob_short:.2f})',
|
| 180 |
+
'direction': 'SHORT',
|
| 181 |
+
'confidence': prob_long, # ✅ إضافة احتمالية الصعود (ستكون منخفضة)
|
| 182 |
+
'short_confidence': prob_short
|
| 183 |
}
|
| 184 |
|
| 185 |
# إذا وصلنا هنا، فالاتجاه هو LONG (شراء)
|
|
|
|
| 189 |
if confidence < CONFIDENCE_THRESHOLD:
|
| 190 |
return {
|
| 191 |
'action': 'WAIT',
|
| 192 |
+
'reason': f'Low Confidence ({confidence:.2f})',
|
| 193 |
+
'direction': 'LONG',
|
| 194 |
+
'confidence': confidence, # ✅ إضافة القيمة للعرض
|
| 195 |
+
'short_confidence': prob_short
|
| 196 |
}
|
| 197 |
|
| 198 |
# 3. التنبؤ بالأهداف والقوة
|
|
|
|
| 209 |
tp_labels = ['TP1', 'TP2', 'TP3', 'TP4']
|
| 210 |
target_profile = tp_labels[tp_class_idx]
|
| 211 |
|
| 212 |
+
# 4. حساب المستويات السعرية
|
| 213 |
atr_pct_val = features['1h_atr_pct'].iloc[0]
|
| 214 |
+
# حماية من القيم الصفرية للـ ATR
|
| 215 |
+
if atr_pct_val == 0: atr_pct_val = 0.02
|
| 216 |
+
|
| 217 |
atr_abs = atr_pct_val * current_price
|
| 218 |
|
| 219 |
tp_map = {
|
|
|
|
| 224 |
}
|
| 225 |
|
| 226 |
primary_tp = tp_map[target_profile]
|
| 227 |
+
sl_price = current_price - (1.2 * atr_abs)
|
| 228 |
|
| 229 |
return {
|
| 230 |
+
'action': 'WATCH',
|
| 231 |
+
'action_type': 'BUY',
|
| 232 |
'confidence': float(confidence),
|
| 233 |
'strength': float(strength),
|
| 234 |
'target_class': target_profile,
|
|
|
|
| 240 |
|
| 241 |
except Exception as e:
|
| 242 |
print(f"❌ [Oracle] Prediction Error: {e}")
|
| 243 |
+
return {'action': 'WAIT', 'reason': 'Error', 'confidence': 0.0}
|