Imputation of Missing Demographic Data in the decide.pe Survey: A First Approach

The dataset was taken from the responses provided by users who completed the decide.pe survey. The responses (n=24851) were submitted between the 14th of march 2026 and the 12th of april 2026. The participants were asked to provide information regarding their geographic location, gender, age, and education. It was not mandatory to respond to the four questions, which led to many gaps in the demographic information. We imputed in a first round the following demographics aspects: age, gender and education. The missing data regarding age was reduced by 4.6% using XGBoost; the best model had an MAE of 7.383, an MSE of 103.466, and an R² of 0.366. The missing data in the gender column was reduced by 2.9%, also using XGBoost; the best model had a macro F1-score of 0.471. Finally, the missing data in the education column was reduced by 4.6% using a Random Forest model, with an F1-score of 0.397. The columns gender, region, age, and education represent the raw data. The columns responses_x_x are the quiz answers. The imputed values are in the columns prediction_x (e.g. prediction_age). The columns x_imputed (e.g. age_imputed) contain the merged values from the raw data and the imputed values. In a second version of this dataset, an imputation of the geographic data is planned.本数据集来源于完成decide.pe问卷调查的用户所提交的回复，有效回复总量为24851条，采集时段为2026年3月14日至2026年4月12日。参与者被要求提供地理位置、性别、年龄与受教育程度相关信息，且上述四项问题均为非必填项，因此人口统计信息存在大量缺失值。我们在第一轮处理中对以下三类人口统计特征开展缺失值填补：年龄、性别及受教育程度。针对年龄字段的缺失值，我们采用XGBoost模型进行填补，使缺失率降低4.6%；最优模型的平均绝对误差（Mean Absolute Error, MAE）为7.383，均方误差（Mean Squared Error, MSE）为103.466，决定系数（R-squared, R²）为0.366。针对性别字段的缺失值，同样采用XGBoost模型进行填补，缺失率降低2.9%；最优模型的宏平均F1值（macro F1-score）为0.471。最后，针对受教育程度字段的缺失值，我们采用随机森林（Random Forest）模型完成填补，缺失率降低4.6%，对应最优模型的F1值为0.397。数据集中，gender（性别）、region（地区）、age（年龄）及education（受教育程度）为原始采集字段；responses_x_x为问卷答题项；填补后的数据存储于prediction_x类字段中（例如prediction_age）；x_imputed类字段（例如age_imputed）则整合了原始数据与填补后的数据。本数据集的第二版计划新增地理信息的缺失值填补工作。