Multi-label Stream Classification Software and Datasets

A Self-Organizing Map for Multi-label Stream Classification

---

Here you can find the source code for the method SOM-AA, as described by Cerri et al., 2022 (To Appear). Our proposal is an online unsupervised incremental method based on self-organizing maps for multi-label stream classification in scenarios with infinitely delayed labels. Differently from Cerri et al., 2021, here we used the algorithm adaptation multi-label approach, training ony one SOM neural network to deal with all classes. All our implementations were performed in R. This README file describes the basic steps to execute our code.

Datasets

---

Our synthetic datasets were constructed using two generatores: one provided with the MOA framework, and one provided by Read et al., 2012. We also adapted two real-world multi-label datasets provided with the Mulan framework. All our datasets can be downloaded from HERE.

Configuration files for the datasets

---

Here you can find the configuration files that can be used to execute the Self-Organizing Maps in the provided datasets. Recall that you can use your own datasets as descrived in the README file. Just construct a corresponding configuration file.