Loading...
Difficulty
Your best score
N/A
You have been recruited as a data scientist in a small "economic prediction lab". Your task is to figure out who breaks the $50,000 per year barrier and who does not, using real data about education, work, and daily life. You have two files with socio-economic data at your disposal, described below:
| Column | Description |
|---|---|
| sampleid | Unique identifier for each observation |
| age | Person's age |
| workclass | Type of employer (e.g., Private, Self-emp-not-inc, etc.) |
| fnlwgt | Final weight from the census |
| education | Educational level (e.g., Bachelors, Masters, etc.) |
| educational_num | Numerical representation of the educational level |
| marital_status | Marital status |
| occupation | Occupation |
| relationship | Relationship in the household |
| race | Race |
| gender | Gender |
| capital_gain | Capital gains |
| capital_loss | Capital losses |
| hours_per_week | Hours worked per week |
| native_country | Country of origin |
| profile_description | Textual description of the professional profile |
| income | Annual income (<=50K / >50K) – target column |
Note:
The income column is only available in the training data (train.csv).
Analyze the training data (train.csv) and determine:
Which country of origin (native_country), other than United-States, has the highest number of people with an income >50K?
The result must be a string with the exact name of the country as it appears in the data.
Analyze the training data (train.csv) and determine:
Which occupation (occupation) has the highest rate of income >50K?
The income rate >50K for an occupation is calculated as: the number of people with income >50K in that occupation, divided by the total number of people in that occupation.
The result must be a string with the exact name of the occupation as it appears in the data.
Build a classification model capable of predicting the value of the income column for all observations in test.csv.
The model must produce one of the two classes:
<=50K>50KThe profile_description column contains textual descriptions of professional profiles. Apply a clustering method on this column to group the observations from test.csv into clusters.
Return the cluster label (integer numeric value) for each observation in test.csv.
Subtask 1: Exact match with the correct answer.
Subtask 2: Exact match with the correct answer.
Subtask 3: Evaluation is performed using F1 Macro:
f1_macro ≥ 0.80 → 55 pointsf1_macro < 0.60 → 0 pointsSubtask 4: Evaluation is performed using Adjusted Rand Index (ARI):
ARI ≥ 0.90 → 20 pointsARI ≤ 0.00 → 0 pointsThe submission file must be in CSV format, with the following columns:
subtaskID – the subtask indexdatapointID – unique identifier from the datasetanswer – the answer or predictionExample:
subtaskID,datapointID,answer1,1,India2,2,Sales3,10001,<=50K3,10002,>50K4,10001,04,10002,2Note:
datapointID must be 1 and 2, respectively.datapointID must be the sampleid value from test.csv.Adult / Census Income Dataset – UCI Machine Learning Repository