Databoost/optimized_adult_census

Documentation of the Dataset - Adult Census Income Dataset Optimized

1. General Description of the Dataset

This dataset, called Adult Census Income Dataset Optimized, is an optimized version of the Adult Census Income Dataset. The latter comes from the UCI Machine Learning Repository and is commonly used in classification tasks to predict whether a person earns more or less than $50,000 per year based on various demographic characteristics.

We optimized the dataset by adding two new columns: capital_gain_loss_ratio and age_group, to enrich the quality of the features. These additions aim to improve model performance, provided they are properly integrated and preprocessed.

Data Sources

The data was extracted from the U.S. Census Survey, a demographic census conducted by the United States Census Bureau. It was used by the Census Bureau to estimate the income of adults based on different personal and social characteristics. The dataset is available on the UCI Machine Learning Repository and is commonly used in machine learning research, particularly for classification tasks.

The dataset contains the following information:

• Demographic data such as age, sex, marital status, occupation, and ethnicity.
• Economic characteristics, such as income (target for classification), hours worked per week, and capital gains/losses.

Purpose of Creation

The primary goal of creating this dataset was to allow studies on the influence of certain socio-economic and demographic characteristics on individuals' income. The dataset was designed for use in machine learning analysis and data studies to predict a person’s income based on these characteristics. The target problem, "predict whether a person earns more than $50,000 or not", is intended for testing supervised classification models.

The intent behind its creation is to provide a case study for researchers, machine learning practitioners, and students to test, train, and evaluate classification models (e.g., logistic regression, decision trees, SVM).

2. AI Use Cases

1. Prediction of Individual Income

The main use case of this dataset is predicting an individual's annual income (more or less than $50,000) based on different characteristics such as age, education, occupation, hours worked per week, and other demographic factors. This problem can be framed as a binary classification (income > 50,000 or not).

Possible Applications:

• Development of credit scoring models: Financial institutions could use this type of model to assess an individual's repayment capacity based on their socio-economic characteristics.
• Analysis of income inequality: AI could be used to explore the factors influencing income gaps in society and identify specific trends related to demographic groups.

2. Evaluation of Economic and Social Inequalities

Researchers can use this dataset to study income disparities between different groups based on criteria such as gender, education, or ethnicity. AI can be used to analyze the correlation between these factors and determine potential biases in the labor market.

Possible Applications:

• Public policy recommendation systems: AI can identify social groups most vulnerable to poverty and recommend targeted policies to improve equal opportunities.
• Economic inequality forecasting: Predictive models can provide insights into the evolution of income gaps based on changes in socio-economic factors.

3. Improving Hiring and Inclusion Strategies

By analyzing this dataset, businesses can implement decision-support systems for human resources. This could include predictive hiring models aimed at identifying candidates most likely to succeed in a given role while considering diversity and inclusion aspects.

Possible Applications:

• Fairness in hiring decisions: Using AI to reduce biases in recruitment, ensuring that factors such as gender, age, or ethnicity do not unduly influence hiring decisions.
• Optimizing compensation policies: Identifying criteria that disproportionately affect compensation to ensure more equitable salary policies.

4. Prediction of the Impact of Education and Training Programs

The dataset contains information on individuals' education levels, allowing the prediction of the impact of educational and training programs on income. AI can be used to analyze the economic benefits of education and professional qualifications.

Possible Applications:

• Cost-benefit analysis of professional training: AI can predict the economic impact of various types of professional training on individuals' future incomes.
• Optimization of educational investments: Governments and businesses can use AI models to invest in the most effective educational programs to maximize economic returns on their investments.

5. Analysis of Demographic Trends in Social Data

AI can be used to extract long-term trends from this dataset, such as the evolution of work behaviors or income changes based on age, gender, and other demographic criteria.

Possible Applications:

• Forecasting labor market trends: AI can help predict future changes in the job market, such as the evolution of the highest-paying jobs.
• Identifying underrepresented populations: Using classification models to detect groups underrepresented in certain professions or sectors.

3. Description of the Columns

Number of Columns: 17 (including the two new columns)

Number of Rows: 32,562 rows

1. age

   • Description: The individual's age in years.
   • Type: int64
   • Use: Age is often used to segment individuals into age groups (e.g., young, adult, senior).

2. workclass

   • Description: The individual's work category. It describes the type of employment status, such as:
     • Private: Private sector employee
     • Self-emp-not-inc: Unincorporated self-employed
     • Self-emp-inc: Incorporated self-employed
     • Federal-gov: Federal government employee
     • Local-gov: Local government employee
     • State-gov: State government employee
     • Without-pay: Unpaid
     • Never-worked: Never worked
   • Type: string

3. fnlwgt

   • Description: The "final weight," a weight used to adjust the sample to properly represent the U.S. population.
   • Type: int64

4. education

   • Description: The individual's education level.
   • Type: string

5. education.num

   • Description: A numerical version of the education level.
   • Type: int64

6. marital.status

   • Description: The marital status of the individual.
   • Type: string

7. occupation

   • Description: The type of profession the individual is engaged in.
   • Type: string

8. relationship

   • Description: The individual's relationship to the household reference person (e.g., spouse, child).
   • Type: string

9. race

   • Description: The individual's racial background.
   • Type: string

10. sex

• Description: The individual's sex.
• Type: string

11. capital.gain

• Description: The capital gains realized by the individual.
• Type: int64

12. capital.loss

• Description: The capital losses suffered by the individual.
• Type: int64

13. hours.per.week

• Description: The number of hours worked per week.
• Type: int64

14. native.country

• Description: The individual's country of origin.
• Type: string

15. age_group

• Description: This column groups individuals into three age groups:
  • Young: Individuals aged 0-30 years.
  • Middle-aged: Individuals aged 30-60 years.
  • Senior: Individuals aged 60-100 years.
• Type: int64
• Purpose: This column allows for population segmentation by age, useful for demographic analyses and income trend studies.

16. capital_gain_loss_ratio

• Description: The ratio between capital gains and capital losses.
• Type: int64
• Purpose: This column helps analyze the relationship between capital gains and losses, useful for financial analysis or studying the impact of investments on income.

17. income

• Description: The individual's annual income. It can be either:
  • ">50K": Income greater than $50,000
  • "<=50K": Income less than or equal to $50,000
• Type: string (target variable for prediction models).