The dataset tracks young South African jobseekers across multiple survey rounds. Each row is one participant's observation in a given round, with their prior-round history summarised inline in the "lag" columns. Rounds 1–8 form the training data; Round 9 is the evaluation cohort you must predict.
# Kaggle_LabourMarketPrediction
## Notes on the data we're working with:
### Target
employed_status — the outcome to predict. 1 = employed, 0 = unemployed. This column appears in train.csv and is removed from test.csv.
### A note on the lag columns
Columns ending in \_lag describe the participant's previous round. They are empty for participants appearing for the first time, who have no earlier round to reference. How you handle this missingness is an important part of the challenge, and note that train and test can differ in how completely some columns are populated.
### Columns
#### Identification and wave timeline
anonymised_id — unique participant identifier and submission key
current_round — the survey round of this observation
lag_round — the round number of the participant's previous observation
days_since_last_obs — days elapsed since the previous observation
total_historical_rounds — number of rounds the participant has appeared in
survey_date — date the observation was recorded
sample — the survey wave/cohort label for this observation
sample_first — the survey wave label of the participant's first appearance
Labour market outcomes and historical lags
employed_lag — employment status in the previous round
status_lag — detailed labour market status in the previous round
status_broad_lag — broad status in the previous round, including a "studying" category
tenure_lag — job tenure (in days) recorded at the previous round
Demographics and geography
age — participant's age at the observation
gender — gender
race — population group
sa_citizen — South African citizen (True/False)
province — province of residence
district — district municipality
municipality — local municipality
Educational attainment
education_level — highest education level attained
education_schooling_grade_twelve_equiv — whether the participant holds a Grade 12 equivalent
institution_type — type of tertiary institution attended (populated …