Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -64,92 +64,91 @@ class EnergyMLPredictor:
|
|
64 |
data = json.loads(json_input)
|
65 |
|
66 |
# Handle both single object and array formats
|
67 |
-
if isinstance(data, list):
|
68 |
-
|
69 |
-
data = data[0]
|
70 |
-
|
71 |
-
# Parse input data
|
72 |
-
date_obj = datetime.strptime(data['data'], '%Y-%m-%d')
|
73 |
-
|
74 |
-
# Color mapping
|
75 |
-
color_mapping = {0: 'incolor', 1: 'verde', 2: 'cinza', 3: 'bronze'}
|
76 |
-
cor_str = color_mapping.get(data['cor'], 'incolor')
|
77 |
-
|
78 |
-
# Create input features (with autoclave)
|
79 |
-
input_data = {
|
80 |
-
'boosting': data['pot_boost'],
|
81 |
-
'espessura': data['espessura'],
|
82 |
-
'extracao_forno': data['extracao_forno'],
|
83 |
-
'porcentagem_caco': data['porcentagem_caco'],
|
84 |
-
'cor': cor_str,
|
85 |
-
'prod_e': data['Prod_E'],
|
86 |
-
'prod_l': data['Prod_L'],
|
87 |
-
'autoclave': data.get('autoclave', 1),
|
88 |
-
'week_day': date_obj.weekday(),
|
89 |
-
'month': date_obj.month,
|
90 |
-
'quarter': (date_obj.month - 1) // 3 + 1,
|
91 |
-
'is_weekend': int(date_obj.weekday() >= 5),
|
92 |
-
'week_of_year': date_obj.isocalendar()[1]
|
93 |
-
}
|
94 |
-
|
95 |
-
# Convert to DataFrame
|
96 |
-
input_df = pd.DataFrame([input_data])
|
97 |
-
|
98 |
-
# Preprocess
|
99 |
-
X_processed = self.threshold_preprocessor.transform(input_df)
|
100 |
-
|
101 |
-
# Make predictions with error handling
|
102 |
-
try:
|
103 |
-
prob_83_raw = self.threshold_model_83.predict_proba(X_processed)
|
104 |
-
prob_83 = prob_83_raw[0][1] if len(prob_83_raw[0]) > 1 else prob_83_raw[0][0]
|
105 |
-
# Ensure probability is between 0 and 1
|
106 |
-
prob_83 = max(0.0, min(1.0, float(prob_83)))
|
107 |
-
except Exception as e:
|
108 |
-
print(f"Error with threshold_83 prediction: {e}")
|
109 |
-
prob_83 = 0.0
|
110 |
-
|
111 |
-
pred_83 = int(prob_83 > 0.5)
|
112 |
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
# Ensure probability is between 0 and 1
|
117 |
-
prob_90 = max(0.0, min(1.0, float(prob_90)))
|
118 |
-
except Exception as e:
|
119 |
-
print(f"Error with threshold_90 prediction: {e}")
|
120 |
-
prob_90 = 0.0
|
121 |
|
122 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
123 |
|
124 |
# Format response
|
125 |
-
next_date = (date_obj + pd.Timedelta(days=1)).strftime('%Y-%m-%d')
|
126 |
-
|
127 |
result = {
|
128 |
"predictions": {
|
129 |
-
"prediction_1":
|
130 |
-
|
131 |
-
"datetime": data['data'],
|
132 |
-
"probabilidade_de_estouro": float(prob_83),
|
133 |
-
"estouro_previsto": pred_83
|
134 |
-
},
|
135 |
-
{
|
136 |
-
"datetime": next_date,
|
137 |
-
"probabilidade_de_estouro": float(prob_83 * 0.98),
|
138 |
-
"estouro_previsto": int(prob_83 * 0.98 > 0.5)
|
139 |
-
}
|
140 |
-
],
|
141 |
-
"prediction_2": [
|
142 |
-
{
|
143 |
-
"datetime": data['data'],
|
144 |
-
"probabilidade_de_estouro": float(prob_90),
|
145 |
-
"estouro_previsto": pred_90
|
146 |
-
},
|
147 |
-
{
|
148 |
-
"datetime": next_date,
|
149 |
-
"probabilidade_de_estouro": float(prob_90 * 0.99),
|
150 |
-
"estouro_previsto": int(prob_90 * 0.99 > 0.5)
|
151 |
-
}
|
152 |
-
]
|
153 |
}
|
154 |
}
|
155 |
|
|
|
64 |
data = json.loads(json_input)
|
65 |
|
66 |
# Handle both single object and array formats
|
67 |
+
if not isinstance(data, list):
|
68 |
+
data = [data]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
+
# Process all items
|
71 |
+
results_83 = []
|
72 |
+
results_90 = []
|
|
|
|
|
|
|
|
|
|
|
73 |
|
74 |
+
for item in data:
|
75 |
+
# Parse input data
|
76 |
+
date_obj = datetime.strptime(item['data'], '%Y-%m-%d')
|
77 |
+
|
78 |
+
# Color mapping
|
79 |
+
color_mapping = {0: 'incolor', 1: 'verde', 2: 'cinza', 3: 'bronze'}
|
80 |
+
if isinstance(item['cor'], str):
|
81 |
+
cor_str = item['cor'].lower()
|
82 |
+
else:
|
83 |
+
cor_str = color_mapping.get(item['cor'], 'incolor')
|
84 |
+
|
85 |
+
# Handle different field name formats for threshold
|
86 |
+
boosting_val = item.get('pot_boost', item.get('ext_boosting', 3.0))
|
87 |
+
|
88 |
+
# Create input features (with autoclave)
|
89 |
+
input_data = {
|
90 |
+
'boosting': boosting_val,
|
91 |
+
'espessura': item['espessura'],
|
92 |
+
'extracao_forno': item['extracao_forno'],
|
93 |
+
'porcentagem_caco': item['porcentagem_caco'],
|
94 |
+
'cor': cor_str,
|
95 |
+
'prod_e': item.get('Prod_E', item.get('prod_e', 1)),
|
96 |
+
'prod_l': item.get('Prod_L', item.get('prod_l', 1)),
|
97 |
+
'autoclave': item.get('autoclave', 1),
|
98 |
+
'week_day': date_obj.weekday(),
|
99 |
+
'month': date_obj.month,
|
100 |
+
'quarter': (date_obj.month - 1) // 3 + 1,
|
101 |
+
'is_weekend': int(date_obj.weekday() >= 5),
|
102 |
+
'week_of_year': date_obj.isocalendar()[1]
|
103 |
+
}
|
104 |
+
|
105 |
+
# Convert to DataFrame
|
106 |
+
input_df = pd.DataFrame([input_data])
|
107 |
+
|
108 |
+
# Preprocess
|
109 |
+
X_processed = self.threshold_preprocessor.transform(input_df)
|
110 |
+
|
111 |
+
# Make predictions with error handling
|
112 |
+
try:
|
113 |
+
prob_83_raw = self.threshold_model_83.predict_proba(X_processed)
|
114 |
+
prob_83 = prob_83_raw[0][1] if len(prob_83_raw[0]) > 1 else prob_83_raw[0][0]
|
115 |
+
# Ensure probability is between 0 and 1
|
116 |
+
prob_83 = max(0.0, min(1.0, float(prob_83)))
|
117 |
+
except Exception as e:
|
118 |
+
print(f"Error with threshold_83 prediction: {e}")
|
119 |
+
prob_83 = 0.0
|
120 |
+
|
121 |
+
pred_83 = int(prob_83 > 0.5)
|
122 |
+
|
123 |
+
try:
|
124 |
+
prob_90_raw = self.threshold_model_90.predict_proba(X_processed)
|
125 |
+
prob_90 = prob_90_raw[0][1] if len(prob_90_raw[0]) > 1 else prob_90_raw[0][0]
|
126 |
+
# Ensure probability is between 0 and 1
|
127 |
+
prob_90 = max(0.0, min(1.0, float(prob_90)))
|
128 |
+
except Exception as e:
|
129 |
+
print(f"Error with threshold_90 prediction: {e}")
|
130 |
+
prob_90 = 0.0
|
131 |
+
|
132 |
+
pred_90 = int(prob_90 > 0.5)
|
133 |
+
|
134 |
+
# Add to results
|
135 |
+
results_83.append({
|
136 |
+
"datetime": item['data'],
|
137 |
+
"probabilidade_de_estouro": round(prob_83, 4),
|
138 |
+
"estouro_previsto": pred_83
|
139 |
+
})
|
140 |
+
|
141 |
+
results_90.append({
|
142 |
+
"datetime": item['data'],
|
143 |
+
"probabilidade_de_estouro": round(prob_90, 4),
|
144 |
+
"estouro_previsto": pred_90
|
145 |
+
})
|
146 |
|
147 |
# Format response
|
|
|
|
|
148 |
result = {
|
149 |
"predictions": {
|
150 |
+
"prediction_1": results_83,
|
151 |
+
"prediction_2": results_90
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
}
|
153 |
}
|
154 |
|