WORLDSOILS - Fully Connected Neural Network

Check the Wiki page for Documentation.

Check the Project tab for tracking tasks.

To download database and Source code/datasets you can use :

git clone https://github.com/saberioon/WORLDSOILS.git

datasets are available in : data folder

To setup environment, first install poetry then using following code in your terminal:

poetry install

To activate your virtual environment use this :

poetry shell

merging_data.py:

merging two csv files based on their ID

python src/merging_data.py clhs_lucas15.csv S2a_resampled_lucas15.csv

fnn.py:

Modelling LUCAS or BSSL data using FNN

python src/fnn.py -i /path/to/input_file -o /path/to/output_folder -l 3 -e 400 -b 20 

-h: help

-l : number of hidden layer ; default = 3

-e : epoch size; default = 400

-b : batch size; default = 20

-d : droup out ; default = 0.2

-r : learning rate ; default = 0.01

-k : kernel initializer ; default = he_normal

LUCAS15_analysis.py:

Modelling LUCAS or BSSL data using FNN

python src/LUCAS15_analysis.py -i /path/to/input_file -o /path/to/output_folder

to see other keys use this :

python src/LUCAS15_analysis.py -h

metrics.py:

calculating R-square, RPD, MSE, RMSE

outlier_finder.py:

Finding outliers using Mahalanobis distance + Principal Component Analysis

Updated: 26 July 2021.