IGSA-RF

IGSA-RF

IGSA-RF: Optimizing Random Forest with Improved Gravitational Search Algorithm

This repository contains the Python implementation of the IGSA-RF algorithm for evaluating the impact of Digital Inclusive Finance on Rural Labor Economics (RLE). It is developed for the research article:

Modeling Rural Labor Responses to Digital Finance: A Hybrid IGSA-Random Forest Approach
Lin, Z, Mathematics, 2025.

Overview

This project uses an Improved Gravitational Search Algorithm (IGSA) to optimize key hyperparameters of a Random Forest model. It then applies the optimized model to estimate Gini and OOB-based feature importance values in a rural labor indicator system.

Key features:

• IGSA implementation for optimizing RF parameters: n_estimators, max_features, min_samples_split
• Gini coefficient and permutation-based OOB importance
• Model performance evaluation (R², MSE, MAE, MAPE)

File Structure

• igsa_rf_model.py: Main script for data processing, optimization, training and evaluation
• 训练数据.xlsx: Input dataset (you must add this file)
• README.md: Project description and instructions

⚙Requirements

• Python 3.8+
• pandas
• numpy
• scikit-learn
• rfpimp (for permutation importances)

Install dependencies via:

pip install pandas numpy scikit-learn rfpimp