Update dataset to v2.0.0: change to native parquet

README.md

@@ -1,97 +1,100 @@
 ---
 license: cc-by-nc-4.0
+tags:
+  - TAAC2026
+  - recommendation
 ---
 
-# TAAC2026 Demo Dataset (1000 Samples)
+# TAAC2026 Demo Dataset (1000 Samples) 
 
-A sample dataset containing 1000 user-item interaction records for the [TAAC2026 competition](https://algo.qq.com).
 
-## Dataset Description
+> [!WARNING] Important Notice
+> **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, rename feature names and update some features.
+> Participants should refer to the updated `demo_1000.parquet` and this `README.md` for the latest schema and data details.
 
-- **Rows**: 1,000
-- **Format**: Parquet (`sample_data.parquet`)
-- **File Size**: ~68 MB
+
+A sample dataset containing 1000 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
+
+| Property | Value |
+|---|---|
+| **File** | `demo_1000.parquet` |
+| **Rows** | 1,000 |
+| **Columns** | 120 |
+| **File Size** | ~39 MB |
 
 ## Columns
 
-| Column | Type | Description |
+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` / `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 |
+
+---
+
+## Detailed Column Schema
+
+### ID & Label Columns (5 columns)
+
+| Column | Arrow Type | Nulls 
 |---|---|---|
-| `item_id` | `int64` | **Target item** identifier. |
-| `item_feature` | `array[struct]` | Array of **target item** feature dicts. Each element has `feature_id`, `feature_value_type`, and value fields (`float_value`, `int_array`, `int_value`). |
-| `label` | `array[struct]` | Array of label dicts. Each element contains `action_time` and `action_type`. |
-| `seq_feature` | `struct` | Sequence features dict with keys: `action_seq`, `content_seq`, `item_seq`. Each sub-key contains arrays of feature structs. |
-| `timestamp` | `int64` | Event timestamp. |
-| `user_feature` | `array[struct]` | Array of user feature dicts. Each element has `feature_id`, `feature_value_type`, and value fields (`float_array`, `int_array`, `int_value`). |
-| `user_id` | `string` | User identifier. |
-
-## Feature Struct Schema
-
-Each feature element contains `feature_id`, `feature_value_type`, and several value fields. Depending on `feature_value_type`, the corresponding value fields are populated and the rest are `null`.
-
-**`item_feature`** — value fields: `int_value`, `float_value`, `int_array`
-
-```json
-{
-  "feature_id": 6,
-  "feature_value_type": "int_value",
-  "float_value": null,
-  "int_array": null,
-  "int_value": 96,
-}
-```
+| `user_id` | `int64` | 0 | 
+| `item_id` | `int64` | 0 | 
+| `label_type` | `int32` | 0 | 
+| `label_time` | `int64` | 0 | 
+| `timestamp` | `int64` | 0 |
 
-**`user_feature`** — value fields: `int_value`, `float_array`, `int_array`
+### User Int Features (46 columns)
 
-```json
-{
-  "feature_id": 65,
-  "feature_value_type": "int_value",
-  "float_array": null,
-  "int_array": null,
-  "int_value": 19
-}
-```
+- `user_int_feats_{1,3,4,48-59,82,86,92-109}`: Scalar `int64`, total 35 columns.
+- `user_int_feats_{15, 60, 62-66, 80, 89-91}`: Array `list<int64>`, total 11 columns.
 
-**`seq_feature`** — value fields: `int_array`
 
-```json
-{
-  "feature_id": 19,
-  "feature_value_type": "int_array",
-  "int_array": [1, 1, 1, ...]
-}
-```
+### User Dense Features (10 columns)
 
-Possible `"feature_value_type"` values and their corresponding fields:
-- `"int_value"` → `int_value` 
-- `"float_value"` → `float_value` 
-- `"int_array"` → `int_array` 
-- `"float_array"` → `float_array`
-- Also there are some combinations of these types, e.g. `"int_array_and_float_array"` → both `int_array` and `float_array` are populated.
+- `user_dense_feats_{61-66, 87, 89-91}`: Array `list<float>`, total 10 columns.
 
-## Label Schema
 
-Each element in the `label` array:
+### Item Int Features (14 columns)
 
-```json
-{
-  "action_time": 1770694299,
-  "action_type": 1
-}
-```
+`item_int_feats_{5-10, 12-13, 16, 81, 83-85}`: Scalar `int64`, total 13 columns.
+
+`item_int_feats_{11}`: Array `list<int64>`, total 1 column.
+
+
+### Domain Sequence Features (45 columns)
+
+`list<int64>` sequences from 4 behavioral domains:
+
+- `domain_a_seq_{38-46}`: 9 columns
+- `domain_b_seq_{67-79, 88}`: 14 columns
+- `domain_c_seq_{27-37, 47}`: 12 columns
+- `domain_d_seq_{17-26}`: 10 columns
+
+---
 
 ## Usage
 
 ```python
+import pyarrow.parquet as pq
 import pandas as pd
 
-df = pd.read_parquet("sample_data.parquet")
-print(df.shape)       # (1000, 7)
-print(df.columns)     # ['item_id', 'item_feature', 'label', 'seq_feature', 'timestamp', 'user_feature', 'user_id']
+# Read the parquet file
+df = pd.read_parquet("demo_1000.parquet")
+
+print(df.shape)       # (1000, 120)
+print(df.columns)     # ['user_id', 'item_id', 'label_type', ...]
 ```
 
 With Hugging Face `datasets`:
-
 ```python
 from datasets import load_dataset
 

sample_data.parquet → demo_1000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3cbea3e8469a5486626cf3e78aeb2c1f4d96ffeb2d5f1683e11529dd1da5bff2
-size 71154363
+oid sha256:0ceaa386ec322436231f2615e5ee89e791d79c727a788c0fb439b0ecb6b68848
+size 40268995