This notebook uses Adult Census Dataset from Kaggle and predict an individual income based on some features

This data was extracted from the 1994 Census bureau database by Ronny Kohavi and Barry Becker (Data Mining and Visualization, Silicon Graphics). A set of reasonably clean records was extracted using the following conditions: ((AAGE>16) && (AGI>100) && (AFNLWGT>1) && (HRSWK>0)). The prediction task is to determine whether a person makes over $50K a year. Data

• Age
• workclass
• fnlwgt
• education
• education.num
• marital.status
• occupation
• relationship
• race
• sex
• capital.gain
• capital.loss
• hours.per.week
• native.country
• Target Values: income

1. Problem Definition: predicting individual income based on provided features
2. Data Exploration:
   • Age distribution was right skewed so we had to handle this
   • fnlwgt distribution was right skewed so also we had to handle this
   • native.country has some rare values that can be combined together
   • relation has some values that can be combined together
   • capital.gain affects individual income although there are many zeros
   • unlike capital.gain, capital.loss doesn't affect the income greatly and has many zeros
   • hours per week affects income.
   • Education & education.num are Two equivilant columns, so we choose eductaion.num to keep.
3. Data Preparation
   • Drop missing values in form of '?' values.
   • Apply log to Age values to make it nearly normally distributed.
   • Apply Log to fnlwgt values to make it nearly normally distributed.
   • Drop Education column
   • Drop capital.loss column
   • change native.country values to be {United-states, Not-Unitied-states}.
   • change ['Married-Spouse-absent', 'Widowed', 'Divorced'] in relation column to be 'Separated'.
   • Encoding Categorical columns in the dataset using pandas function get_dummies.
   • split data to Train set and Test set with ratios 70%, 30% respectively.
4. Model Training
   • model : GradientBoostingClassifier with default values but with max_depth = 5.
5. Model Evaluation
   • Accuracy: 86.09%