Redescription mining augmented with random forest of multi-target predictive clustering trees

Mihelčić, Matej; Džeroski, Sašo; Lavrač, Nada; Šmuc, Tomislav

doi:10.1007/s10844-017-0448-5

Cited by 16 publications

(16 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In case maximal size of the redescription set is reached, the algorithm exchanges the newly produced redescription with the worst incomplete candidate from R (lines 11 to 19 in Algorithm 1 [28]. In this work we use three models: a) The random forest of multi-target regression (multi-label classification) PCTs [18], b) The Random Forest of Extra randomized multi-target PCTs [17] and c) The Random Forest of multi-target regression PCTs with Random Output Selections [5] (see Section IV-D for motivation and more details).…”

Section: A the Gclus-rm Algorithmmentioning

confidence: 99%

“…The model M obtained using algorithm Alg is used to generate targets that connect two views and is transformed to rules used to create redescriptions (more detailed explanation can be seen in [26]). Rules obtained from a model M , obtained using Alg , are used to increase diversity and accuracy of produced redescriptions [28]. In this work, we only test Predictive Clustering trees and Extra multi-target PCTs (M) as models from which main rules are generated.…”

Section: A Preliminariesmentioning

confidence: 99%

“…In addition to providing thorough results of a framework's performance and comparison to the naive way of creating multi-view redescription mining algorithms by utilizing existing two-view approaches, we provide experiments of using supplementing forest of Predictive Clustering trees (PCTs) [28], with [5] (PCT-ROS) and without [18] random output selection and using a supplementing forest of Extra multi-target Predictive Clustering trees [17] (EPCT) to increase the overall performance. Using a supplementing model obtained by training a Random Forest of PCTs was shown to increase accuracy and diversity of produced redescriptions when two views are used [28]. In this work, we test the same hypothesis on datasets using more than two views and test supplementing models produced by training PCTs with random output selection and using random forest of Extra multi-target Predictive Clustering trees.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Approaches for Multi-View Redescription Mining

Mihelčić

Šmuc

2021

IEEE Access

View full text Add to dashboard Cite

The task of redescription mining explores ways to re-describe different subsets of entities contained in a dataset and to reveal non-trivial associations between different subsets of attributes, called views. This interesting and challenging task is encountered in different scientific fields, and is addressed by a number of approaches that obtain redescriptions and allow for the exploration and analyses of attribute associations. The main limitation of existing approaches to this task is their inability to use more than two views. Our work alleviates this drawback. We present a memory efficient, extensible multi-view redescription mining framework that can be used to relate multiple, i.e. more than two views, disjoint sets of attributes describing one set of entities. The framework can use any multi-target regression or multi-label classification algorithm, with models that can be represented as sets of rules, to generate redescriptions. Multi-view redescriptions are built using incremental view-extending heuristic from initially created twoview redescriptions. In this work, we use different types of Predictive Clustering trees algorithms (regular, extra, with random output selection) and the Random Forest thereof in order to improve the quality of final redescription sets and/or execution time needed to generate them. We provide multiple performance analyses of the proposed framework and compare it against the naive approach to multi-view redescription mining. We demonstrate the usefulness of the proposed multi-view extension on several datasets, including a use-case on understanding of machine learning models-a topic of growing importance in machine learning and artificial intelligence in general. INDEX TERMS knowledge discovery, multi-view redescription mining, redescription set, predictive clustering trees, random forest, extremely randomized trees, random output selection.

show abstract

Section: A the Gclus-rm Algorithmmentioning

confidence: 99%

Section: A Preliminariesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Approaches for Multi-View Redescription Mining

Mihelčić

Šmuc

2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The RF is an ensemble learning algorithm [20] extended from DT algorithm and trained by the ensemble learning theory of bagging. In RF algorithm, once the tree classifier is established, each DT will receive a result about the class of a newly inputted user sample.…”

Section: Random Forest (Rf) Algorithmmentioning

confidence: 99%

A Big Data-Based Anti-Fraud Model for Internet Finance

Liu¹,

You²

2020

RIA

View full text Add to dashboard Cite

Fraud has become a serious risk to the burgeoning industry of Internet finance. To predict and prevent the fraud, this paper develops a risk control model that accurately predicts the fraud behaviors of online loan borrowers, based on their social network data and consumption information. Firstly, statistical analysis was carried out to clarify the relationship between user behavior and fraud, and that between social network and fraud. Next, suitable variables were selected through big data analysis, and used to build the risk control model, in combination with random forest (RF) algorithm and modern financial theory. The effectiveness of the proposed model was confirmed through contrastive experiments. Our model provides peer-to-peer (P2P) online loan platforms with an effective tool to prevent fraud.

show abstract

“…Attribute-based explo- Table 2 Algorithm parameters used to create redescription set on a Country dataset (C), DBLP dataset (D) and Phenotype dataset (P). 2 × 40 denotes two random initialization with 40 iterations each and 1 + R F 100 denotes using a PCT augmented with Random Forest of 100 PCTs (Mihelcic et al 2018) ration context might require changes given specificities of a given task (e.g associations in association rule mining are one sided and directed). InterSet can also facilitate understanding of supervised problems (in particular multi-label predictive problems), if redescription mining is performed on input attributes and target attributes as two separate views.…”

Section: Specifics and Motivation For Developing Intersetmentioning

confidence: 99%

Targeted and contextual redescription set exploration

Mihelčić

Šmuc

2018

Mach Learn

Self Cite

View full text Add to dashboard Cite

One important problem occurring in redescription mining is a very large number of produced redescriptions. This makes analyses time consuming and generally difficult. We present the targeted and contextual redescription set exploration, realized through the tool InterSet. The main purpose of the tool is to derive additional knowledge from the redescription set which allows exploring parts of redescription set of interest and examining redescriptions individually or in the broader context, with the aim of increasing overall understandability. InterSet allows relating, grouping redescriptions, observing distributions of various redescription properties and selecting the appropriate subsets for further, detailed study. This allows gaining knowledge about the underlying data, help in forming, understanding, supporting research hypothesis or assists in understanding one or more redescriptions of interest. The tool provides three different, fully connected interaction modes based on: (1) similarity of entity occurrence in redescription support sets, (2) attribute co-occurrence in redescriptions and (3) redescription quality measures. Additionally, it allows exploration of relations between different redescriptions by creating a graph visualization that includes the top k-shortest paths containing selected redescriptions. On the individual redescription level, it allows studying value distributions of described entities, for a given set of attributes.

show abstract

Redescription mining augmented with random forest of multi-target predictive clustering trees

Cited by 16 publications

References 21 publications

Approaches for Multi-View Redescription Mining

Approaches for Multi-View Redescription Mining

A Big Data-Based Anti-Fraud Model for Internet Finance

Targeted and contextual redescription set exploration

Contact Info

Product

Resources

About