GENII

This is the Pytorch implementation of GEMII in the paper: [GENII: A Graph Neural Network-based Model for Citywide Litter Prediction Leveraging Crowdsensing Data].

The baseline's code and model code are all in the folder code. the data preprocessing code and chart drawing code can be found in the folder code_preliminary for reference.

The trained city model parameters and baseline data are in the folder data_base, and the processed dataset can be downloaded at here.

It is worth noting that, as delineated in the paper, the outcomes generated by our code when executed on disparate machines may exhibit minor variations attributable to distinct machine parameters and related factors.

Requirements

• pytorch

• pytorch-geometric

• scikit-learn

• python 3

• jupyter notebook

Usage

To run the code

1. Need to download the processed dataset firstly at here.
2. cd to code_model : Execute dataset.py, Krommenie_dataset.py, and Hyattsville_dataset.py to individually generate graph datasets for the three cities.
3. cd to code_model : python model_train.py or python Krommenie_model_train.py or python Hyattsville_model_train.py.
4. To launch the graphical user interface, run interface.py