Association Rule Mining (ARM) for plant species distribution.
This is the repository associated with the paper entitled "Application of association rule mining to assess forest species distribution in Italy considering abiotic and biotic factors", co-authored by Valeria Aloisi, Sergio Noce, Italo Epicoco, Cristina Cipriano, Massimo Cafaro, Giuseppe Brundu, Lorenzo Arcidiaco, Donatella Spano, Giovanni Aloisio, and Simone Mereu.
The study aim was to explore the relationships that may exist among co-occurring forest species, considering both biotic and abiotic conditions, and focusing on 151 plant species monitored at 6,784 sites across Italy, thus improving the assessment of forest community distribution.
The repository contains the Python code developed for association rule mining, applied to a synthetically generated dataset, including 396 sites and 11 possible forest species. Unfortunately, the real dataset exploited in the research study included data available only upon access consent by the "Comando Unità Forestale, Ambientale e Agroalimentare (CUFAA) of the Arma dei Carabinieri". Therefore, the association rules resulting from the execution of this code on the synthetic dataset are different from those obtained and reported in the aforementioned research paper.
The Python script named "ARM_code_on_forest_species_synthetic_data.py" was executed using Python 3.8.2 and requires the following packages:
- Pandas 1.0.3
- NumPy 1.19.5
- mlxtend 0.23.1
The script can be verified by using the "synthetic_dataset.csv" as the input dataset. The output is a CSV file containing the resulting association rules with the three key metrics: support, confidence, and lift. Specifically, "synthetic_dataset.csv" is a sample file used to verify the correct installation of the application.
The application can be executed through the following CLI:
$> python3 ARM_code_on_forest_species_synthetic_data.py