Update README.md

README.md

@@ -43,7 +43,7 @@ pretty_name: Beverage Energy Tracker
 size_categories:
 - 10K<n<100K
 ---
-# Dataset Card for Beverage Energy Tracker
+# 📑 Dataset Card for Beverage Energy Tracker
 
 This dataset card documents the **Beverage Energy Tracker** dataset.  
 It contains nutritional and energy-related features of 30 unique beverages, with an augmented split expanding to 300+ rows.  
@@ -65,6 +65,7 @@ It contains nutritional and energy-related features of 30 unique beverages, with
 **Direct Use**  
 - Educational practice in data collection, preprocessing, and augmentation.  
 - Demonstration of label-preserving jitter augmentation for tabular datasets.  
+- Binary classification task: predict whether a beverage is **high sugar** or not.  
 
 **Out-of-Scope Use**  
 - Not intended for clinical, nutritional, or health policy decision-making.  
@@ -75,25 +76,27 @@ It contains nutritional and energy-related features of 30 unique beverages, with
 ## Dataset Structure
 
 - **Original split:** 30 manually curated beverages.  
-- **Augmented split:** 300 rows, generated with Gaussian jitter (for continuous features) and ±1 step encoding (for ordinal features).  
+- **Augmented split:** 300+ rows, generated with Gaussian jitter (for continuous features) and ±1 step encoding (for ordinal features).  
 
 **Features:**  
 - `Beverage type` (string)  
 - `Added sugar (g)` (float)  
 - `Calories` (float)  
 - `Volume (mL)` (integer)  
-- `Energy rating (1–5)` (ordinal target label)  
+- `Energy rating (1–5)` (ordinal)  
+- `is_high_sugar` (binary target: `1 = sugar ≥ 15g`, `0 = sugar < 15g`)  
 
 ---
 
 ## Dataset Creation
 
 **Curation Rationale**  
-To study how nutritional features (sugar, calories, volume) can relate to an *energy rating*.  
+To study how nutritional features (sugar, calories, volume) can relate to **sugar content classification** (high vs low).  
 
 **Data Collection and Processing**  
 - Data manually collected from: Starbucks, ALDI USA, MarketDistrict, Amazon, Walmart product pages.  
 - Manual preprocessing and type formatting were applied (e.g., ensuring numeric columns, clipping noise to avoid negative values).  
+- Binary target `is_high_sugar` was derived using a threshold of **15g sugar**.  
 
 **Source Data Producers**  
 - Original producers: beverage manufacturers and retailers (nutritional info posted publicly).  
@@ -103,7 +106,7 @@ To study how nutritional features (sugar, calories, volume) can relate to an *en
 
 ## Annotations
 
-- **Annotation Process:** Energy rating was assigned manually (scale 1–5).  
+- **Annotation Process:** Binary target derived from numeric sugar values.  
 - **Annotators:** Dataset creator.  
 
 ---
@@ -117,7 +120,7 @@ To study how nutritional features (sugar, calories, volume) can relate to an *en
 ## Bias, Risks, and Limitations
 
 - Limited to 30 beverages (not representative of all products).  
-- Energy rating is subjective, not an industry standard.  
+- `Energy rating` and `is_high_sugar` are simplified labels, not standardized nutrition metrics.  
 - Augmentation may create unrealistic numeric combinations.  
 
 ---