Invasive Weed Optimization (IWO) Algorithm Feature Selection

• Invasive Weed Optimization (IWO) algorithm feature selection
• You can use your own data or change number of features by "nf"
• "nVar" should be equal with total number of features

---

This repository implements the Invasive Weed Optimization (IWO) algorithm for feature selection and evaluates its performance using a Neural Network-based cost function. The project includes loading datasets, feature selection, and comparison of classification performance before and after feature selection using K-Nearest Neighbors (KNN).

---

Features

• Dataset Loading: Load input features and target values from a dataset (iwodata).
• Feature Selection: Optimize feature selection using the IWO algorithm.
• Artificial Neural Network (ANN): Train and evaluate a neural network to calculate feature selection cost.
• K-Nearest Neighbors (KNN): Compare classification accuracy on the original dataset and selected features.
• Performance Visualization: Iteration-wise cost tracking and confusion matrices for evaluation.

---

Usage

1. Dataset: Place your dataset file (iwodata.mat) in the same directory as the scripts. The dataset should include:

   • Inputs: Features matrix.
   • Targets: Labels matrix.

2. Run the main script: Execute the IWO Feature Selection.m file in MATLAB.

3. Results:

   • Best-selected features are stored in BestSol.out.S.
   • Compare KNN performance before and after feature selection.

---

Key Scripts

1. Loading.m:

   • Loads the dataset and extracts inputs and targets.

2. CreateAndTrainANN.m:

   • Creates and trains a neural network using the Levenberg-Marquardt algorithm.
   • Divides data into training, validation, and testing subsets.

3. FeatureSelectionCost.m:

   • Defines the cost function for IWO.
   • Evaluates the performance of selected features using ANN.

4. IWO Feature Selection.m:

   • Implements the Invasive Weed Optimization algorithm.
   • Compares KNN accuracy before and after feature selection.
   • Visualizes performance metrics.

---

Dependencies

• MATLAB R2019b or later.
• Neural Network Toolbox.

---

• Quote: "If you set your goals ridiculously high and it's a failure, you will fail above everyone else's success.

James Cameron