sourander/yskills

---
license: cc-by-nc-4.0
language:
- en
---

## Data Source

Original data has been downloaded from Data in Brief article dataset [Digital skills among youth: A dataset from a three-wave longitudinal survey in six European countries](https://www.sciencedirect.com/science/article/pii/S2352340924003652)

## Dataset Processing Summary

Modifications Applied:

1. Wave Consolidation: The original wide-format data (separate columns for waves 1-3) was converted to long format, concatenating all three waves into rows.
2. Target Variable: Only rows with valid RISK101 values (Experience with cyberhate in the past year) were retained.
3. Feature Selection: Cherry-picked 22 derived/aggregated columns plus 3 demographic columns (country, Age_year, GENDER), avoiding multicollinear features (e.g., kept lit_inf_pro but dropped skill_inf_pro).
4. Missing Value Handling: Negative values (representing various "not available" codes in ySKILLS) were replaced with NaN and imputed using mode (most frequent value).
5. Encoding & Scaling:

* Categorical columns (country, Age_year, GENDER): One-hot encoded (with first category dropped)
* Ordinal/continuous columns: Standardized using StandardScaler
* Binary columns (skill_progr, RISK101): Left unchanged

### Research Task

Binary classification: Predict `RISK101` (cyberhate experience) from digital skills and demographic features.

许可证：CC BY-NC 4.0（知识共享署名-非商业性使用4.0国际许可协议）
语言：英语

---

### 数据来源
原始数据下载自《Data in Brief》期刊论文数据集《Digital skills among youth: A dataset from a three-wave longitudinal survey in six European countries》（https://www.sciencedirect.com/science/article/pii/S2352340924003652）

### 数据集处理概述
所应用的修改如下：
1. **波次整合**：将原始宽格式数据（为3个调查波次分别设置独立列）转换为长格式，将三个波次的所有数据合并至行维度。
2. **目标变量筛选**：仅保留`RISK101`（过去一年遭遇网络仇恨言论经历）取值有效的样本行。
3. **特征选择**：精选22个衍生/聚合特征与3个人口统计特征（国家、Age_year、GENDER），规避多重共线性特征（例如保留`lit_inf_pro`，移除`skill_inf_pro`）。
4. **缺失值处理**：将ySKILLS数据集中代表各类“不可用”编码的负值替换为NaN，并通过众数（出现频率最高的值）进行插补。
5. **编码与标准化**：
* 分类特征（国家、Age_year、GENDER）：采用独热编码（删除首个类别以避免虚拟变量陷阱）
* 有序/连续特征：使用StandardScaler进行标准化处理
* 二分类特征（`skill_progr`、`RISK101`）：保持原样未作修改

### 研究任务
二分类任务：基于数字技能与人口统计特征预测`RISK101`（网络仇恨言论遭遇经历）。