Added more performance metrics to README

README.md

@@ -82,7 +82,7 @@ The code used to train the model can be found on GitHub:
 
 The research paper can be found here: [ELECTRA and GPT-4o: Cost-Effective Partners for Sentiment Analysis](https://github.com/jbeno/sentiment/research_paper.pdf)
 
-### Performance
+### Performance Summary
 
 - **Merged Dataset**
     - Macro Average F1: **79.29**
@@ -97,7 +97,6 @@ The research paper can be found here: [ELECTRA and GPT-4o: Cost-Effective Partne
     - Macro Average F1: **69.95**
     - Accuracy: **78.24**
 
-
 ## Model Architecture
 
 - **Base Model**: ELECTRA base discriminator (`google/electra-base-discriminator`)
@@ -255,6 +254,112 @@ The model's configuration (config.json) includes custom parameters:
 - `dropout_rate`: Dropout rate used in the classifier.
 - `pooling`: Pooling strategy used ('mean').
 
+## Performance by Dataset
+
+### Merged Dataset
+
+```
+Merged Dataset Classification Report
+
+              precision    recall  f1-score   support
+
+    negative   0.847081  0.777211  0.810643      2352
+     neutral   0.704453  0.761072  0.731669      1829
+    positive   0.828047  0.844615  0.836249      2349
+
+    accuracy                       0.796937      6530
+   macro avg   0.793194  0.794299  0.792854      6530
+weighted avg   0.800285  0.796937  0.797734      6530
+
+ROC AUC: 0.926344
+
+Predicted  negative  neutral  positive
+Actual                                
+negative       1828      331       193
+neutral         218     1392       219
+positive        112      253      1984
+
+Macro F1 Score: 0.79
+```
+
+### DynaSent Round 1
+
+```
+DynaSent Round 1 Classification Report
+
+              precision    recall  f1-score   support
+
+    negative   0.901222  0.737500  0.811182      1200
+     neutral   0.745957  0.922500  0.824888      1200
+    positive   0.850970  0.804167  0.826907      1200
+
+    accuracy                       0.821389      3600
+   macro avg   0.832716  0.821389  0.820992      3600
+weighted avg   0.832716  0.821389  0.820992      3600
+
+ROC AUC: 0.945131
+
+Predicted  negative  neutral  positive
+Actual                                
+negative        885      201       114
+neutral          38     1107        55
+positive         59      176       965
+
+Macro F1 Score: 0.82
+```
+
+### DynaSent Round 2
+
+```
+DynaSent Round 2 Classification Report
+
+              precision    recall  f1-score   support
+
+    negative   0.696154  0.754167  0.724000       240
+     neutral   0.770408  0.629167  0.692661       240
+    positive   0.704545  0.775000  0.738095       240
+
+    accuracy                       0.719444       720
+   macro avg   0.723702  0.719444  0.718252       720
+weighted avg   0.723702  0.719444  0.718252       720
+
+ROC AUC: 0.88842
+
+Predicted  negative  neutral  positive
+Actual                                
+negative        181       26        33
+neutral          44      151        45
+positive         35       19       186
+
+Macro F1 Score: 0.72
+```
+
+### Stanford Sentiment Treebank (SST-3)
+
+```
+SST-3 Classification Report
+
+              precision    recall  f1-score   support
+
+    negative   0.831878  0.835526  0.833698       912
+     neutral   0.452703  0.344473  0.391241       389
+    positive   0.834669  0.916392  0.873623       909
+
+    accuracy                       0.782353      2210
+   macro avg   0.706417  0.698797  0.699521      2210
+weighted avg   0.766284  0.782353  0.772239      2210
+
+ROC AUC: 0.885009
+
+Predicted  negative  neutral  positive
+Actual                                
+negative        762      104        46
+neutral         136      134       119
+positive         18       58       833
+
+Macro F1 Score: 0.70
+```
+
 ## License
 
 This model is licensed under the MIT License.