rtnjt commited on
Commit
627ce6d
·
verified ·
1 Parent(s): 748ae82

updated app.py again

Browse files
Files changed (1) hide show
  1. app.py +64 -19
app.py CHANGED
@@ -14,41 +14,86 @@ models = {
14
  "AdaBoost": joblib.load('adaboost_model.joblib')
15
  }
16
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17
  def predict_risk(model_name, gender, age, social_category, domicile_type, first_gen,
18
  disability_status, income_bracket, family_members, earning_members,
19
  scholarship, loan_status, class10, class12, cgpa, backlogs, year_backs):
20
 
21
- # Select the model based on user choice
22
  model = models[model_name]
23
 
24
- # Create input dataframe matching training data format
25
- # Create input dataframe matching training data format
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26
  input_data = pd.DataFrame({
27
- 'Gender_bin': [gender],
28
  'Age': [age],
29
- 'Social_Category': [social_category], # Check if this needs a suffix too
30
- 'Domicile_Type': [domicile_type],
31
- 'First_Gen_bin': [first_gen],
32
- 'Disability_Status_bin': [disability_status],
33
- 'Income_ord': [income_bracket],
34
  'Total_Family_Members': [family_members],
35
  'Earning_Members': [earning_members],
36
- 'Scholarship_Status_bin': [scholarship],
37
- 'Education_Loan_Status_bin': [loan_status],
38
- 'Class10_ord': [class10],
39
- 'Class12_ord': [class12],
40
  'Current_CGPA': [cgpa],
41
  'Current_Backlogs': [backlogs],
42
  'Year_Backs': [year_backs]
43
  })
44
 
45
- # Make prediction
46
  prediction = model.predict(input_data)[0]
47
  result = "At Risk" if prediction >= 2 else "Not At Risk"
48
 
49
- # Calculate confidence
50
- probabilities = model.predict_proba(input_data) if hasattr(model, 'predict_proba') else None
51
- confidence = float(np.max(probabilities)) * 100 if probabilities is not None else 85.0
 
 
 
52
 
53
  return {
54
  "model_used": model_name,
@@ -57,7 +102,7 @@ def predict_risk(model_name, gender, age, social_category, domicile_type, first_
57
  "risk_score": float(prediction)
58
  }
59
 
60
- # Create Gradio interface
61
  iface = gr.Interface(
62
  fn=predict_risk,
63
  inputs=[
@@ -89,4 +134,4 @@ iface = gr.Interface(
89
  )
90
 
91
  if __name__ == "__main__":
92
- iface.launch()
 
14
  "AdaBoost": joblib.load('adaboost_model.joblib')
15
  }
16
 
17
+ # --- BACKGROUND MAPPINGS ---
18
+ # These dictionaries map the "Friendly Text" -> "Model Number"
19
+ # You MUST ensure these numbers match exactly what your LabelEncoder/OneHotEncoder used during training.
20
+ mappings = {
21
+ "Gender": {"Male": 1, "Female": 0},
22
+ "YesNo": {"Yes": 1, "No": 0},
23
+ "Income": {
24
+ "Below ₹3,00,000": 0,
25
+ "₹3,00,001 - ₹5,00,000": 1,
26
+ "₹5,00,001 - ₹8,00,000": 2,
27
+ "₹8,00,001 - ₹12,00,000": 3,
28
+ "Above ₹12,00,000": 4
29
+ },
30
+ "Grades": {
31
+ "Below 50%": 0,
32
+ "50% - 59.9%": 1,
33
+ "60% - 69.9%": 2,
34
+ "70% - 79.9%": 3,
35
+ "80% - 89.9%": 4,
36
+ "90% and above": 5
37
+ },
38
+ # Add mappings for Social Category and Domicile if your model needs numbers
39
+ "Social_Category": {"General": 0, "OBC": 1, "SC": 2, "ST": 3, "EWS": 4},
40
+ "Domicile_Type": {"Urban": 0, "Semi-Urban": 1, "Rural": 2}
41
+ }
42
+
43
  def predict_risk(model_name, gender, age, social_category, domicile_type, first_gen,
44
  disability_status, income_bracket, family_members, earning_members,
45
  scholarship, loan_status, class10, class12, cgpa, backlogs, year_backs):
46
 
47
+ # 1. Select the model
48
  model = models[model_name]
49
 
50
+ # 2. SILENT CONVERSION (Text -> Number)
51
+ # We use .get() to safely grab the number. Default to 0 if something goes wrong.
52
+ encoded_gender = mappings["Gender"].get(gender, 0)
53
+ encoded_first_gen = mappings["YesNo"].get(first_gen, 0)
54
+ encoded_disability = mappings["YesNo"].get(disability_status, 0)
55
+ encoded_scholarship = mappings["YesNo"].get(scholarship, 0)
56
+ encoded_loan = mappings["YesNo"].get(loan_status, 0)
57
+ encoded_income = mappings["Income"].get(income_bracket, 0)
58
+ encoded_class10 = mappings["Grades"].get(class10, 0)
59
+ encoded_class12 = mappings["Grades"].get(class12, 0)
60
+
61
+ # Handle the ones that might have caused the "could not convert string" error
62
+ # If your model used OneHotEncoder for these, you might need a different approach,
63
+ # but if it used LabelEncoder, this is perfect.
64
+ encoded_social = mappings["Social_Category"].get(social_category, 0)
65
+ encoded_domicile = mappings["Domicile_Type"].get(domicile_type, 0)
66
+
67
+ # 3. Create the numeric DataFrame
68
  input_data = pd.DataFrame({
69
+ 'Gender_bin': [encoded_gender],
70
  'Age': [age],
71
+ 'Social_Category': [encoded_social],
72
+ 'Domicile_Type': [encoded_domicile],
73
+ 'First_Gen_bin': [encoded_first_gen],
74
+ 'Disability_Status_bin': [encoded_disability],
75
+ 'Income_ord': [encoded_income],
76
  'Total_Family_Members': [family_members],
77
  'Earning_Members': [earning_members],
78
+ 'Scholarship_Status_bin': [encoded_scholarship],
79
+ 'Education_Loan_Status_bin': [encoded_loan],
80
+ 'Class10_ord': [encoded_class10],
81
+ 'Class12_ord': [encoded_class12],
82
  'Current_CGPA': [cgpa],
83
  'Current_Backlogs': [backlogs],
84
  'Year_Backs': [year_backs]
85
  })
86
 
87
+ # 4. Predict
88
  prediction = model.predict(input_data)[0]
89
  result = "At Risk" if prediction >= 2 else "Not At Risk"
90
 
91
+ # 5. Confidence
92
+ try:
93
+ probabilities = model.predict_proba(input_data)
94
+ confidence = float(np.max(probabilities)) * 100
95
+ except:
96
+ confidence = 85.0
97
 
98
  return {
99
  "model_used": model_name,
 
102
  "risk_score": float(prediction)
103
  }
104
 
105
+ # Create Gradio interface (Users see Text, Code sees Numbers)
106
  iface = gr.Interface(
107
  fn=predict_risk,
108
  inputs=[
 
134
  )
135
 
136
  if __name__ == "__main__":
137
+ iface.launch()