Farmer Income Prediction using Random Forest Regression

This project aims to predict the annual income of farmers using demographic, socio-economic, and environmental features. The model is built using Random Forest Regression, with appropriate data cleaning, feature engineering, and evaluation steps.

---

Dataset

• Provided train and test datasets.
• Each record represents information about a farmer and their environment.
• The target variable is Income (transformed using log for modeling).

---

Key Steps & Highlights

1. Exploratory Data Analysis (EDA)

• Checked distribution of target variable (Income)
• Identified positive skew and applied log transformation
• Analyzed categorical variables like SEX, REGION, MARITAL_STATUS
• Visualized outliers and distributions

2. Data Preprocessing

• Encoded categorical variables (e.g., SEX, REGION, etc.)
• Dropped less useful or problematic categorical columns with object dtype

3. Feature Scaling

• Identified numerical columns with more than 2 unique values
• Applied StandardScaler to normalize continuous features
• Retained one-hot encoded or binary features without scaling

4. Model Building

• Used Linear Regression from sklearn
• Trained model on log-transformed income (Income_Log)
• Ensured consistent column order and structure in test set

5. Model Evaluation

• Calculated MAPE (Mean Absolute Percentage Error) on training data
• Achieved a MAPE of ~21.06% (~78.94% average prediction accuracy)

6. Final Prediction

• Reversed the log transformation to get predicted income
• Generated final CSV/Excel file with Farmer_ID and predicted income

---

Technologies Used

• Python (Pandas, NumPy, Scikit-learn)
• Matplotlib / Seaborn
• Git & GitHub for version control

---