Flood Probability Prediction using Neural Networks

This project is based on a Kaggle competition Playground Series - Season 4, Episode 5 that focuses on predicting flood probability. The model is implemented using TensorFlow and Keras, with an emphasis on fine-tuning for performance.

Project Files

• train.csv: The training dataset containing features and flood probability labels.
• test.csv: The test dataset used for generating predictions.
• sample_submission.csv: A sample submission file provided by Kaggle to format the prediction results.
• main.ipynb: The Jupyter Notebook from which this preprocessing, training, evaluating and creating submission file.
• submission.csv: The final submission file generated by the model.

Project Overview

The goal of this project is to predict the probability of flooding based on various environmental and infrastructural factors. The model uses a Sequential neural network with dense layers, trained on the provided dataset. The project involves data preprocessing, model training, and evaluation.

Data Preprocessing

• Combined training and test datasets for consistent preprocessing.
• Created new features Ratio1 and Ratio2 derived from the existing features.
• Scaled the data using RobustScaler to handle outliers and ensure a uniform scale across features.

Model Architecture

• Dense Layer 1: 64 units, ReLU activation.
• Dense Layer 2: 32 units, ReLU activation.
• Output Layer: 1 unit, linear activation for regression.

The model is compiled with the Adam optimizer and a mean squared error loss function. A custom R^2 score metric is also used to evaluate model performance.

Model Training

• Batch Size: 1024
• Epochs: 100
• The model is trained on the training set, with evaluation on the same set due to the absence of a validation split in this script.

Model Evaluation

The model's performance is evaluated using the R^2 score and loss on the training set. The best model based on these metrics is saved and used for generating predictions on the test set.

Installation

To run this project, ensure you have Python installed along with the required libraries listed in requirements.txt.

pip install -r requirements.txt

Usage

1. Run the Training Script On:

   main.ipynb

   This will train the model and save the predictions in submission.csv.

2. Submit Predictions: Upload submission.csv to the Kaggle competition to evaluate the model's performance.

License

This project is made by Hüseyin Battal and is intended for educational purposes as part of the Kaggle competition.

GitHub: https://github.com/huseyinbattal3469
LinkedIn: https://www.linkedin.com/in/huseyin-battal/