Update dataset to v2.0.0: use native parquet format

README.md

@@ -1,5 +1,5 @@
 ---
-license: apache-2.0
+license: cc-by-nc-4.0
 task_categories:
   - text-classification
 tags:
@@ -9,11 +9,15 @@ size_categories:
   - n<1K
 ---
 
-# TAAC2026 Demo Dataset — data_1000
+# TAAC2026 Demo Dataset 
 
-> ⚠️ **Warning**: This is a demo dataset for demonstration purposes only. It is not intended for model training or evaluation, and may not reflect the full complexity of the actual TAAC2026 data.
+> ⚠️ **Update[2026.04.10]**: This demo dataset has been updated to newest version with the following changes:
+> - The parquet file is now a **flat column layout**, with all features as top-level columns.
+> - Add a sequence feature and update some user/item features.
+> Participants should refer to the updated `demo_1000.parquet` and this README for the latest schema and data details.
 
-A sample dataset containing **1,000** user-item interaction records for the TAAC2026 competition. This dataset uses a **flat column layout** — all features are stored as individual top-level columns instead of nested structs/arrays. It is a 1,000-row subset (first 1,000 rows) of the `demo_1000_0408.gz.parquet` (1,016 rows).
+
+A sample dataset containing **1,000** user-item interaction records for [the TAAC2026 competition](https://algo.qq.com/). This dataset uses a **flat column layout** — all features are stored as individual top-level columns instead of nested structs/arrays.
 
 ## Dataset Overview
 
@@ -22,11 +26,8 @@ A sample dataset containing **1,000** user-item interaction records for the TAAC
 | **File** | `demo_1000.parquet` |
 | **Rows** | 1,000 |
 | **Columns** | 120 |
-| **Format** | Apache Parquet (uncompressed) |
-| **File Size** | ~38.41 MB |
-| **Unique Users** | 1,000 |
-| **Unique Items** | 837 |
-| **Timestamp Range** | `2026-03-05 23:36:40` — `2026-03-05 23:49:41` |
+| **Format** | Apache Parquet |
+| **File Size** | ~38.38 MB |
 
 ## Label Distribution
 
@@ -42,9 +43,9 @@ The 120 columns fall into **6 categories**:
 | Category | Count | Arrow Type | Description |
 |---|---|---|---|
 | **ID & Label** | 5 | `int64` / `int32` | Core identifiers, label, and timestamp |
-| **User Int Features** | 46 | `int64` / `double` / `list<int64>` | Integer-valued user features (scalar or array) |
-| **User Dense Features** | 10 | `list<float>` | Float-array user features (e.g. embeddings) |
-| **Item Int Features** | 14 | `double` / `list<int64>` | Integer-valued item features (scalar or array) |
+| **User Int Features** | 46 | `int64` / `list<int64>` | Integer-valued user features (scalar or array) |
+| **User Dense Features** | 10 | `list<float>` | Float-array user features |
+| **Item Int Features** | 14 | `int64` / `list<int64>` | Integer-valued item features (scalar or array) |
 | **Domain Sequence Features** | 45 | `list<int64>` | Behavioral sequence features from 4 domains |
 
 ---
@@ -63,7 +64,7 @@ The 120 columns fall into **6 categories**:
 
 ### User Int Features (46 columns)
 
-#### Scalar Columns (`int64` / `double`)
+#### Scalar Columns (`int64`)
 
 | Column | Nulls | Null% | Min | Max | Mean | Unique |
 |---|---|---|---|---|---|---|
@@ -140,20 +141,20 @@ All columns are `list<float>` arrays (e.g. embedding vectors).
 
 | Column | Arrow Type | Nulls | Null% | Min | Max | Mean | Unique |
 |---|---|---|---|---|---|---|---|
-| `item_int_feats_5` | `double` | 2 | 0.2% | 4 | 325 | 118.452 | 82 |
-| `item_int_feats_6` | `double` | 2 | 0.2% | 0 | 977 | 419.073 | 216 |
-| `item_int_feats_7` | `double` | 2 | 0.2% | 0 | 2,806 | 1,052.866 | 349 |
-| `item_int_feats_8` | `double` | 2 | 0.2% | -1 | 2,431 | 463.712 | 226 |
-| `item_int_feats_9` | `double` | 2 | 0.2% | 3 | 37 | 21.171 | 24 |
-| `item_int_feats_10` | `double` | 2 | 0.2% | 2 | 309 | 150.007 | 110 |
+| `item_int_feats_5` | `int64` | 2 | 0.2% | 4 | 325 | 118.452 | 82 |
+| `item_int_feats_6` | `int64` | 2 | 0.2% | 0 | 977 | 419.073 | 216 |
+| `item_int_feats_7` | `int64` | 2 | 0.2% | 0 | 2,806 | 1,052.866 | 349 |
+| `item_int_feats_8` | `int64` | 2 | 0.2% | -1 | 2,431 | 463.712 | 226 |
+| `item_int_feats_9` | `int64` | 2 | 0.2% | 3 | 37 | 21.171 | 24 |
+| `item_int_feats_10` | `int64` | 2 | 0.2% | 2 | 309 | 150.007 | 110 |
 | `item_int_feats_11` | `list<int64>` | 439 | 43.9% | — | — | — | — |
-| `item_int_feats_12` | `double` | 2 | 0.2% | 0 | 2,777 | 1,039.381 | 352 |
-| `item_int_feats_13` | `double` | 2 | 0.2% | 1 | 8 | 4.457 | 8 |
-| `item_int_feats_16` | `double` | 2 | 0.2% | 2 | 35,259 | 12,356.101 | 662 |
-| `item_int_feats_81` | `double` | 2 | 0.2% | 0 | 2 | 0.508 | 3 |
-| `item_int_feats_83` | `double` | 832 | 83.2% | 1 | 31 | 17.595 | 22 |
-| `item_int_feats_84` | `double` | 832 | 83.2% | 3 | 226 | 131.131 | 66 |
-| `item_int_feats_85` | `double` | 832 | 83.2% | 4 | 1,001 | 439.816 | 103 |
+| `item_int_feats_12` | `int64` | 2 | 0.2% | 0 | 2,777 | 1,039.381 | 352 |
+| `item_int_feats_13` | `int64` | 2 | 0.2% | 1 | 8 | 4.457 | 8 |
+| `item_int_feats_16` | `int64` | 2 | 0.2% | 2 | 35,259 | 12,356.101 | 662 |
+| `item_int_feats_81` | `int64` | 2 | 0.2% | 0 | 2 | 0.508 | 3 |
+| `item_int_feats_83` | `int64` | 832 | 83.2% | 1 | 31 | 17.595 | 22 |
+| `item_int_feats_84` | `int64` | 832 | 83.2% | 3 | 226 | 131.131 | 66 |
+| `item_int_feats_85` | `int64` | 832 | 83.2% | 4 | 1,001 | 439.816 | 103 |
 
 ### Domain Sequence Features (45 columns)
 
@@ -251,7 +252,7 @@ print(null_pct[null_pct > 0.5])  # Columns with >50% nulls
 
 ## Key Notes
 
-1. **Schema type difference**: Scalar integer features are stored as `double` (not `int64`) in this parquet file due to the presence of null values (pandas converts nullable int to float).
+1. **Nullable int64**: All `*_int_feats_*` scalar columns are stored as Arrow `int64` with native null support. When reading with pandas, nullable int columns may be converted to `float64` — use `df[col].fillna(-1).astype(int)` or read with `pd.Int64Dtype()` to preserve the integer type.
 2. **No nested structs**: Unlike the older `sample_data.parquet`, all features are flat top-level columns.
 3. **Sparse features**: 18 columns have >40% null values — handle missing data carefully during feature engineering.
 4. **Sequence lengths vary widely**: Domain sequences range from length 1 to ~3,951, which may require truncation or padding for model input.

demo_1000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ada35e38e8728009b8579182b9e53f4ae3ff135f73eca7eecd62c259d7ea45e
-size 40279261
+oid sha256:0ceaa386ec322436231f2615e5ee89e791d79c727a788c0fb439b0ecb6b68848
+size 40268995