arkodeep commited on
Commit
0a71f1f
·
verified ·
1 Parent(s): 41c8a38

Upload 6 files

Browse files
Files changed (6) hide show
  1. metrics.txt +2 -0
  2. model.pkl +3 -0
  3. model_optimized.pkl +3 -0
  4. models.md +37 -0
  5. vectorizer.pkl +3 -0
  6. vectorizer_optimized.pkl +3 -0
metrics.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ Accuracy: 0.9758
2
+ Precision: 1.0000
model.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f13dc88836745131a01a3882112fc4e1b5cc21c93c8559579e577cc691a3a330
3
+ size 24989568
model_optimized.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b273536c5f5e0fc0f532393afec5ec9a972e14970756c4db828235c222127622
3
+ size 29159702
models.md ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Comparison of Training Models
2
+
3
+ 1. **Lite Model**
4
+ - Uses basic ensemble voting with fixed weights [2,1,2]
5
+ - Pre-configured hyperparameters (no optimization)
6
+ - Lemmatization for text preprocessing
7
+ - Uses 3 algorithms: SVC, MultinomialNB, ExtraTreesClassifier
8
+ - Fastest because no parameter tuning/optimization
9
+
10
+ 2. **Legacy Model**
11
+ - Uses simple voting ensemble without weight optimization
12
+ - Porter Stemming for text preprocessing (simpler than lemmatization)
13
+ - Slightly different hyperparameters:
14
+ - SVC with 'sigmoid' kernel
15
+ - Fewer trees in ExtraTreesClassifier (50 vs 200)
16
+ - Medium speed due to simpler preprocessing
17
+
18
+ 3. **Monarch Butterfly Optimization (MBO) Model**
19
+ - Uses nature-inspired optimization algorithm
20
+ - Optimizes 7 parameters simultaneously:
21
+ - SVC parameters (C, gamma)
22
+ - MultinomialNB alpha
23
+ - Number of trees
24
+ - Ensemble weights (w1, w2, w3)
25
+ - Population-based search with:
26
+ - 20 butterflies
27
+ - 30 iterations
28
+ - Cross-validation for each evaluation
29
+ - Slowest because:
30
+ - Runs multiple training cycles (20 butterflies × 30 iterations)
31
+ - Each evaluation requires 5-fold cross-validation
32
+ - Total of ~3000 model evaluations
33
+
34
+ **Summary**:
35
+ - Lite: Quick, fixed parameters
36
+ - Legacy: Traditional, basic ensemble
37
+ - MBO: Advanced optimization, but computationally intensive
vectorizer.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82b35da250caa705d592c11d661a79f20496f04caa5080c9956fd3de5491a19a
3
+ size 105868
vectorizer_optimized.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82af4b40b45ed7a05dd55d80ba6c063200f797cb7331d7a960c5caa2965316a8
3
+ size 197799