Update README.md

README.md

@@ -6,7 +6,7 @@ tags:
 - toxicology
 pretty_name: AttentiveSkin
 dataset_summary: >-
- They compiled GHS dataset comprising 731 Corr., 1283 Irrit., and 1205 Neg. samples 
+ They compiled GHS dataset comprising 731 Corrostion, 1283 Irritation, and 1205 Negative samples 
  from 6 governmental databases and 2 external datasets.
 citation: >-
   @article{,
@@ -41,7 +41,12 @@ dataset_info:
     - name: "CAS RN"
       dtype: string
     - name: "GHS"
-      dtype: string
+      dtype:
+        class_label:
+          names:
+            Cat 1: "Corrosion"
+            Cat 2: "Irritation"
+            NC: "Negative" 
     - name: "Detailed Page"
       dtype: string
     - name: "Evidence"
@@ -141,20 +146,20 @@ then load, featurize, split, fit, and evaluate the catboost model
       representations = load_representations_from_dicts([{"name": "morgan"}, {"name": "maccs_rdkit"}]))
 
     model = load_model_from_dict({
-        "name": "cat_boost_regressor",
+        "name": "cat_boost_classifier",
         "config": {
             "x_features": ['SMILES::morgan', 'SMILES::maccs_rdkit'],
-            "y_features": ['log_solubility'],
-        }})
-        
+            "y_features": ['GHS']}})
+    
     model.train(split_featurised_dataset["train"])
     preds = model.predict(split_featurised_dataset["test"])
     
-    regression_suite = load_suite("regression")
+    classification_suite = load_suite("classification")
     
-    scores = regression_suite.compute(
-        references=split_featurised_dataset["test"]['Solubility'],
-        predictions=preds["cat_boost_regressor::Solubility"])
+    scores = classification_suite.compute(
+        references=split_featurised_dataset["test"]['GHS'],
+        predictions=preds["cat_boost_classifier::GHS"])
+        
 
 ## AttentiveSkin
 To Predict Skin Corrosion/Irritation Potentials of Chemicals via Explainable Machine Learning Methods