💧 River Discharge Forecasting using Machine Learning and AOA Optimization

An intelligent framework for river discharge forecasting using ensemble regression models optimized by the Arithmetic Optimization Algorithm (AOA).
The dataset is prepared with lag features to forecast next-day discharge, but the framework can be easily adapted to other time series forecasting tasks.

---

🧠 Overview

This project integrates machine learning regression models (GBR, SVR, KNN, Voting Regressor)
with the Arithmetic Optimization Algorithm (AOA) to optimize model parameters and ensemble weights.
It is designed to predict tomorrow’s river discharge based on historical flow data, but users can replace the dataset to apply it to any regression or forecasting problem.

The framework automatically:

1. Loads and preprocesses data
2. Splits into training, validation, and test sets
3. Trains multiple regression models
4. Optimizes ensemble weights and hyperparameters using AOA
5. Evaluates models using multiple metrics (R², RMSE, MAE, MSE)
6. Produces visual outputs and exports all results

---

📊 Example Dataset

The provided example uses river discharge data, but users can replace it with any numerical dataset.
To use your own data:

1. Place your file in the ./data/ directory.
2. Update the filename and sheet names in config/setting.yaml.
3. Run the main script again.

---

⚙️ Configuration (config/setting.yaml)

All parameters can be controlled from the configuration file.

data:
  name: "station1"
  path: "./data/station1.xlsx"
  sheet_range: [1, 5]

train:
  split:
    train_end: 2577
    valid_end: 3105
  batch_size: 32
  epochs: 50

optimization:
  method: "AOA"
  population_size: 100
  epochs: 100
  alpha: 5
  miu: 0.5
  moa_min: 0.2
  moa_max: 0.9

results:
  metrics_dir: "./results/metrics/"
  predictions_dir: "./results/predictions/"
  plots_dir: "./results/prediction_plots/"

---

🧩 Workflow

Workflow Summary:

1. Data Loading & Normalization using MinMaxScaler
2. Base Regressors: Gradient Boosting, SVR, KNN
3. Ensemble: Voting Regressor
4. Optimization: AOA for weight and hyperparameter tuning
5. Evaluation: Compute R², RMSE, MAE, MSE
6. Visualization: Scatter plots, violin plots, and time-series comparisons

---

📈 Visual Outputs

Type	Description	Example
Scatter Plots	Predicted vs True values for all models
Violin Plot	Distribution of model predictions
Time Series	Comparison of predicted and true values over time

---

🧾 Results

All results (metrics, predictions, and visualizations) are automatically saved to:

results/
│
├── metrics/             # R², RMSE, MAE, MSE summary
├── predictions/         # Excel files with model outputs
└── prediction_plots/    # All generated plots

Each file is named according to the dataset and scenario defined in the configuration file.

---

📦 Installation

git clone https://github.com/rasoulameri/River_Discharge_Regression.git
cd River_Discharge_Regression
pip install -r requirements.txt

---

▶️ Usage

To run the framework:

python main.py

or for Jupyter visualization:

jupyter notebook notebooks/EDA_and_Visualization.ipynb

---

🧠 Example Models

Model	Description
GBR	Gradient Boosting Regressor
SVR	Support Vector Regressor
KNN	k-Nearest Neighbors Regressor
VR	Ensemble Voting Regressor
AOA_VR	AOA-tuned ensemble with optimal hyperparameters

---

📄 Data Availability Statement

The raw data used in this project (river discharge samples) are provided for demonstration. Users are encouraged to replace them with their own datasets.

---

🏗 Repository Structure

River_Discharge_Regression/
│
├── config/
│   └── setting.yaml                # Main configuration
│
├── data/
│   └── River_Discharge.xlsx        # Example dataset
│
├── src/
│   ├── models.py
│   ├── data_loader.py
│   ├── evaluate.py
│   ├── optimization.py
│   └── train.py
│
├── results/
│   ├── metrics/
│   ├── predictions/
│   └── prediction_plots/
│
├── requirements.txt
│
└─ main.py

---

📫 Contact

Rasoul Ameri
📧 rasoulameri90@gmail.com
🔗 GitHub Profile

---