cissagatto Goto Github PK

graphmultilabel

This code is part of my Ph.D. research. The aim is generate label co-ocurrence graphs from similarity matrices

hpml

This repository hold all experiments conducted during my PhD (2019-2023). HPML means "Hybrid Partitions for Multi-Label Classification". SET-UP-1

hpml-chains

This code is a part of my doctoral research at PPG-CC/DC/UFSCar in colaboration with Ku Leuven in Belgium.

hpml-j

This code is a part of my doctoral research at PPG-CC/DC/UFSCar. HPML-J is the name of the first experiment carried out: Hybrid Partitions for Multi-Label Classification with index Jaccard.

hpml-kais

Repository for the paper "Multi-Label Classification with Label Clusters"

hpml.d.ce

This code is part of my Ph.D. research. The objective is to test the best chosen hybrid partitions with silhouette coefficient. It's called HPML Clusters Chains because we chain the labels of each cluster with subsequent clusters. It is a version of HPML where there is only the external chaining.

hpml.d.cei

This code is part of my Ph.D. research. The objective is to test the best chosen hybrid partitions with silhouette coefficient. A version HPML where both internal and external chaining is performed. This is a joint version of Label Chains HPML and Cluster Chains HPML. Therefore, there is the chaining of labels and clusters.

hpml.d.ci

This code is part of my Ph.D. research. The objective is to test the best chosen hybrid partitions with silhouette coefficient. It's called HPML Label Chains because we use Ensemble of Classifier Chains to test, this means that the labels are randomly chained within each cluster.

hpml.d.padrao

This code is part of my Ph.D. research. The objective is to test the best chosen hybrid partitions with silhouette coefficient. A version HPML where both internal and external chaining is performed. This is the original version of HPML, i.e., a version without any type of chaining.

jaccard

This code generate partitions for a multilabel dataset using the Jaccard Index similarity measure. We use HCLUST with 6 linkage metrics to generate several partitions. You may build the partition with the highest coefficient. This code also provide an analysis about the partitioning.

k-nn-case-study

This repository is a result of my studies about k-NN. Theory and Pratice are describe in this study.

lh_cd_elainegatto

INDICIUM - Processo Seletivo - Lighthouse Programa De Formação Em Dados - Remoto

local-partitions

This code is part of my PhD research. The aim is built and validate local partitions for multi-label classification

mips32bitscheatsheet

Cheatsheet completinha do MIPS 32 bits - MIPS Technologies

mips32bitsexemplocompleto

Um exemplo completinho de Exemplo MIPS: programa principal, funções, vetores, arrays, if, for, while, recursividade

multi-label-friedman-nemenyi

Compute Friedman and Nemenyi Statistical Tests for Multilabel Classification

multilabelperformanceevaluationmeasures

This code shows how to compute the measures of multi-label classification hand in hand.

multilabelsimilaritiesmeasures

Compute similarities measures (categorical data) for all labels in label space for a multilabel dataset

oracle-clus

This code is part of my doctoral research. It's oracle experimentation of Bell Partitions using the CLUS framework.

pair-comparison

One-to-one comparison. Count how many datasets your method (algorithm) obtained the best result when compared to other method (or methods) in your experiment.

propriedades-multirrotulo

r-clus-hmc

This code executes the CLUS algorithm in an R script.

r_multilabel_csv2arff

This code converts CSV file in an ARFF file correctly for a multi-label dataset

random3-clus

rogers

This code generate partitions for a multilabel dataset using the Rogers-Tanimoto similarity measure. We use HCLUST with 6 linkage metrics to generate several partitions. You may build the partition with the highest coefficient. This code also provide an analysis about the partitioning.

scmamp

Statistical comparison of multiple algorithms

similaritiesmultilabel

This code is part of my Ph.D. research. The aim is generate similarity matrices from similarity measures.

tcp-knn-h-clus

Test the best hybrid partition generated by hierarchical community detection methods wiht k-NN sparsification using Clus Framework

tcp-knn-nh-clus

Test the best hybrid partition generated by non hierarchical comunity detection methods, and k-NN sparsification, using Clus Framework.

tcp-random-h-clus

Test the best random partition generated by hierarchical community detection methods using clus framework