Stable Decision Trees: Using Local Anarchy for Efficient Incremental Learning

Papagelis

2009

Soft Comput

Self Cite

When genetic algorithms are used to evolve decision trees, key tree quality parameters can be recursively computed and re-used across generations of partially similar decision trees. Simply storing instance indices at leaves is sufficient for fitness to be piecewise computed in a lossless fashion. We show the derivation of the (substantial) expected speedup on two bounding case problems and trace the attractive property of lossless fitness inheritance to the divide-and-conquer nature of decision trees. The theoretical results are supported by experimental evidence.

Section: Discussionmentioning

confidence: 99%

Lossless fitness inheritance in genetic algorithms for decision trees

Papagelis

2009

Soft Comput

Self Cite

“…To achieve hiding by minimally modifying the original dataset, we may interpret "minimally" concerning dataset changes or whether the sanitized decision tree produced via hiding is syntactically close to the original. Measuring minimality in how one modifies decision trees has been studied regarding heuristics that guarantee or approximate the impact of changes [19][20][21].…”

Section: Methodsmentioning

confidence: 99%

On Using Linear Diophantine Equations for in-Parallel Hiding of Decision Tree Rules

Feretzakis

Verykios

2019

Entropy

Data sharing among organizations has become an increasingly common procedure in several areas such as advertising, marketing, electronic commerce, banking, and insurance sectors. However, any organization will most likely try to keep some patterns as hidden as possible once it shares its datasets with others. This paper focuses on preserving the privacy of sensitive patterns when inducing decision trees. We adopt a record augmentation approach to hide critical classification rules in binary datasets. Such a hiding methodology is preferred over other heuristic solutions like output perturbation or cryptographic techniques, which limit the usability of the data, since the raw data itself is readily available for public use. We propose a look ahead technique using linear Diophantine equations to add the appropriate number of instances while maintaining the initial entropy of the nodes. This method can be used to hide one or more decision tree rules optimally.

“…To achieve hiding by minimally modifying the original data set, we can interpret “minimal” changes in data sets or whether the sanitized decision tree generated through hiding is syntactically close to the original with minimum modification of the initial data set. The minimum measurement of how decision-making trees are changed has been examined in the context of heuristics to ensure or approximate the effect of changes [ 22 , 23 , 24 ]. In our examples, we use the measure of the kappa statistic to compare the efficiency of the deduced decision tree after the proposed modification with the original one.…”

Section: Methodsmentioning

confidence: 99%

Using Minimum Local Distortion to Hide Decision Tree Rules

Feretzakis

Verykios

2019

Entropy

The sharing of data among organizations has become an increasingly common procedure in several areas like banking, electronic commerce, advertising, marketing, health, and insurance sectors. However, any organization will most likely try to keep some patterns hidden once it shares its datasets with others. This article focuses on preserving the privacy of sensitive patterns when inducing decision trees. We propose a heuristic approach that can be used to hide a certain rule which can be inferred from the derivation of a binary decision tree. This hiding method is preferred over other heuristic solutions like output perturbation or cryptographic techniques—which limit the usability of the data—since the raw data itself is readily available for public use. This method can be used to hide decision tree rules with a minimum impact on all other rules derived.