ECS_171_Project_FQ-24

This project aims to predict rainfall in California using three machine learning models: XGBoost, SARIMA, and Random Forest.

Project Workflow

1. Data Preprocessing:

   • Clean and preprocess the input datasets (final_data.csv and Rainfall_Dataset.csv) using scripts in the Scripts directory.
   • Feature engineering and exploratory data analysis are conducted to prepare the data for modeling.

2. Model Development:

   • Train and evaluate three models:
     • XGBoost: Developed and evaluated in xgb_model.ipynb.
     • SARIMA: Implemented in time_series_model.py.
     • Random Forest: Code integrated in the respective Python scripts (3 different files).

3. Forecasting:

   • Generate rainfall forecasts at the county level (County_Forecasts/) and the station level (Station_Forecasts/).

4. Visualization and Metrics:

   • Generate visualizations and calculate evaluation metrics in notebooks like barplot_metrics.ipynb and metrics_df.ipynb.

Models Used

XGBoost

• A powerful gradient boosting algorithm designed for structured data. Used to generate accurate forecasts at both county and station levels.

SARIMA

• A statistical time series forecasting method that extends ARIMA by accounting for seasonality.

Random Forest

• An ensemble learning method based on decision trees. Useful for making robust predictions on complex datasets.

Installation and Usage

Prerequisites

Ensure you have Python installed. Clone the repository and navigate to the project directory.

Install Dependencies

Install required Python libraries using requirements.txt:

pip install -r requirements.txt

Run the Project

1. Preprocessing and Training:
   • Use the Jupyter notebooks in the Scripts/ directory to preprocess data and train the models.
2. Forecasting:
   • Run the .py files in the Scripts/ directory to generate forecasts for counties and stations using different models.

Results

The forecast results are saved in the County_Forecasts/ and Station_Forecasts/ directories as text files.