=Paper=
{{Paper
|id=None
|storemode=property
|title=Classification Methods Based on Formal Concept Analysis
|pdfUrl=https://ceur-ws.org/Vol-977/paper11.pdf
|volume=Vol-977
}}
==Classification Methods Based on Formal Concept Analysis==
https://ceur-ws.org/Vol-977/paper11.pdf
Classification Methods Based on Formal
Concept Analysis
Olga Prokasheva, Alina Onishchenko, and Sergey Gurov
Faculty of Computational Mathematics and Cybernetics, Moscow State University
Abstract. Formal Concept Analysis (FCA) provides mathematical mod-
els for many domains of computer science, such as classification, cate-
gorization, text mining, knowledge management, software development,
bioinformatics, etc. These models are based on the mathematical proper-
ties of concept lattices. The complexity of generating a concept lattice
puts a constraint to the applicability of software systems. In this pa-
per we report on some attempts to evaluate simple FCA-based classifica-
tion algorithms. We present an experimental study of several benchmark
datasets using FCA-based approaches. We discuss difficulties we encoun-
tered and make some suggestions concerning concept-based classification
algorithms.
Keywords: Classification, pattern recognition, data mining, formal con-
cept analysis, biclustering
1 Introduction
Supervised classification consists in building a classifier from a set of examples
labeled by their classes or precedents (learning step) and then predicting the
class of new examples by using the generated classifiers (classification step).
Document classification is a sub-field of information retrieval. Documents may
be classified according to their subjects or according to other attributes (such as
document type, author, year of publication, etc.). Mostly, document classification
algorithms are based on supervised classification. Algorithms of this kind can be
used for text mining, automatical spam-filtering, language identification, genre
classification, text mining. Some modern data mining methods can be naturally
described in terms of lattices of closed sets, i.e., concept lattices [1], also called
Galois lattices. An important feature of FCA-based classification methods do not
make any assumptions regarding statistical models of a dataset. Biclustering [9,
10] is an approach related to FCA: it proposes models and methods alternative
to classical clustering approaches, being based on object similarity expressed
by common sets of attributes. There are several FCA-based models for data
analysis and knowledge processing, including classification based on learning
from positive and negative examples [1, 2].
In our previous work [15] the efficiency of a simple FCA-based binary classifi-
cation algorithm was investigated. We tested this method on different problems
96 Olga Prokasheva, Alina Onishchenko, Sergey Gurov
with numerical data and found some difficulties in its application. The main pur-
pose of this paper is to investigate critical areas of the FCA method for better
understanding of its features. Several advices for developers are also provided.
We test hypothesis-based classification algorithm and our modified FCA-based
method on 8 benchmarks. We describe our experiments and compare the per-
formance of FCA-based algorithms with that of SVM-classification [16].
2 Definitions
Formal Concept Analysis. In what follows we keep to standard FCA defini-
tions from [1]. Let G and M be sets, called set of objects and set of attributes,
respectively. Let I ⊆ G × M be a binary relation. The triple K = (G, M, I) is
called a formal context. For arbitrary A ⊆ G and B ⊆ M the mapping (.)0 is
defined as follows:
A0 = {m ∈ M | ∀g ∈ A(gIm)}; B 0 = {g ∈ G | ∀m ∈ B(gIm)}. (1)
This pair of mappings defines a Galois connection between the sets 2G and
2M partially ordered by the set-theoretic inclusion. Double application of the
operation (·)0 is a closure operator on the union of the sets 2G and 2M . Let a
context K be given. A pair of subsets (A, B), such that A ⊆ G, B ⊆ M , A0 = B,
and B 0 = A is called a formal concept of K with formal extent A and formal
intent B. The extent and the intent of a formal concept are closed sets.
FCA in learning and classification. Here we keep to definitions from [2]
and [3]. Let K = (G, M, I) be a context and w 6∈ M a target attribute.
In FCA terms, the input data for classification may be described by three
contexts w.r.t. w: the positive context K+ = (G+ , M, I+ ), the negative con-
text K− = (G− , M, I− ) , and the undefined context Kτ = (Gτ , M, Iτ ) [2].
G− , G+ and Gτ are sets of positive, negative and undefined objects respectively,
I ⊆ G × M , where ∈ {−, +, τ } are binary relations that define structural
attributes. Operators (·)0 in these contexts are denoted by A+ , A− , Aτ , respec-
tively. For short we write g 0 , g 00 , g + , g − , g τ instead of {g}0 , {g}00 , {g}+ , {g}− , {g}τ ,
respectively. A formal concept of a positive context is called a positive concept.
Negative and undefined concepts, as well as extents and intents of the contexts
K− and Kτ , are defined similarly. A positive formal intent B+ of (A+ , B+ ) ∈ K+
is called a positive or (+) — prehypothesis if is not the formal intent of any neg-
ative concept, and it is called a positive or (minimal) (+) — hypothesis if it is
not a subset of the intent g − for some elementary concept (g, g − ) for a negative
example g; otherwise it is called a false (+)-generalization.
Negative (or (-) —) prehypotheses, hypotheses, and false generalizations are
defined similarly. The definitions imply that a hypothesis is also a prehypothesis.
Hypotheses are used to classify undefined examples from the set Gτ . If unclas-
sified object g τ contains a positive but no negative hypothesis, it is classified
positively, similar for negative. No classification happens if the formal intent g τ
does not contain any subsets of either positive or negative hypotheses (insuffi-
cient data) or contains both a positive and a negative hypothesis (inconsistent
data).
Classification Methods Based on Formal Concept Analysis 97
Biclustering. The particular case of biclustering [10–12] we have considered is
a development of the FCA-based classification method. Using FCA methods, we
can construct a hierarchical structure of biclusters that reflects the taxonomy
of data. Density of bicluster (A, B) of the formal context K = (G, M, I) is
defined as ρ(A, B) = |I ∩ {A × B}|/(|A| · |B|). Specify some value ρmin ∈ [0, 1].
The bicluster (A, B) is called dense if ρ(A, B) ≥ ρmin . Stability index σ of a
concept (A, B) is given by σ(A, B) = |C(A, B)|/2|A| , where C(A, B) is the union
of the sets C ⊆ A such that C = B 0 [13, 21]. Biclusters, as well as dense and
stable formal concepts (i.e., concepts having stability above a fixed threshold),
are used to generate hypotheses for clustering problems [13].
3 Basic Classification Algorithms
Several FCA-based classification methods are known [19, 15]: GRAND [31, 17],
LEGAL [26], GALOIS [25], RULELEARNER [24], CIBLe [30], CLNN&CLNB [27],
NAVIGALA[28], CITREC [29, 17] and classification method based on hypothe-
ses [8, 7, 2, 3]. There are several categories of FCA-based classification methods:
1. Hypothesis-based classification using the general principle described in
Section 2.
2. Concept lattice based classification. A concept lattice can be seen as
a search-space in which one can easily pass from a level to another one.
The navigation can e.g. start from the top concept with the least intent.
Then one can progress concept by concept by taking new attributes and re-
ducing the set of objects. Many systems use lattice-based classification, such
as GRAND [31, 17], RULEARNER [24], GALOIS [25], NAVIGALA [28] and
CITREC [29, 17]. The common constraint of these systems is the exponential
algorithmic complexity of generating a lattice. For this reason, some systems
search in a subset of the set of all concepts.
3. Classification based on Galois sub-hierarchies. Systems like CLNN&
CLNB [27], LEGAL [26] and CIBLe [30] build Galois sub-hierarchy (ordered
set of object and attribute concepts), which drastically reduces algorithmic
complexity.
4. Cover-based classification. A concept cover is a part of the lattice con-
taining only pertinent concepts. The construction of a cover concept is based
on heuristic algorithms which reduce the complexity of learning. The con-
cepts are extracted one by one. Each concept is given by a local optimization
of a measure function that defines pertinent concepts. IPR (Induction of
Product Rules) [32] was the first method generating a concept cover. Each
pertinent concept induced by IPR is given by a local optimization of en-
tropy function. The sets of pertinent generated concepts are sorted from the
more pertinent to the less pertinent and each pertinent concept with the
associated class gives a classification rule.
98 Olga Prokasheva, Alina Onishchenko, Sergey Gurov
4 Classification Experiments with Benchmarks
4.1 A Hypothesis-Based Algorithm
The method for constructing concept-based hypotheses described above inspired
the following binary classification algorithm [15]. The main steps of the algorithm
are as follows:
1. Data binarization. The situation where the attributes are non-binary, but
a classification method is designed for binary data brings up the problem
of attribute binarization, or scaling. This problem is very difficult and a
lot of papers are devoted to it. Scaling problem arises also when we use
FCA for object classification. For specific tasks scaling is usually carried out
empirically by repeatedly solving the problem of classification on precedents.
It is clear, however, that in a couple of ”scaling–recognition method” the
determining factor is exactly scaling. Indeed, in the event of its successful
application a ’good’ transformation of the feature space will be obtained
and almost any recognition algorithm will show good results in that space.
So that the problem of scaling is a nonspecific for FCA-recognition methods
and the current level of development of these methods unable to point the
best technique of scaling focused on their use. That is why our work is not
focused on this problem and we use a simple scaling, which, we believe,
allowed more clearly to identify the features of FCA-classification methods.
Hence, we just normalized all attributes to [0,1] interval and than applied
interval-based nominal scaling. The number of intervals is fixed and equals
10. The size of intervals is also fixed and equals 0.1.
2. Hypothesis generation and classification. Algorithm searches common
attributes for all objects from the first class (second class), which are not
observed for any objects from the second class (first class). Obtained sets of
attributes (hypothesis) are used to classify undefined objects.
The algorithm has been tested on numerical benchmarks. The data for the
first four problems is taken from the UCI Machine Learning Repository1 . Prob-
lem 5 (Two Norm) involving the separation of two normal 20-dimensional dis-
tributions is taken from the University of Toronto site2 ; the CART classification
algorithm [22] produced for this problem an error rate of 22.1% with a train-
ing sample of 300 precedents, which is almost a factor of 10 higher than the
theoretical minimum for the ideal classifier — the Fisher discriminant func-
tion. Problems 6 (Lung Cancer), 7 (Cirrhosis), and 8 (Cloud Seeding) are taken
from the StatLib site3 . The considered problems come from different specific
research areas. For example, in the Liver Disorders problem, the objects are
datasets obtained from the tests of six patients. The training sample consists
of 345 precedents divided into positive and negative classes with respect to the
1
http://archive.ics.uci.edu/ml
2
http://www.cs.toronto.edu/∼delve/data/twonorm/desc.html
3
http://lib.stat.cmu.edu/datasets, pages /veteran, /pbc, and /cloud, respectively
Classification Methods Based on Formal Concept Analysis 99
target attribute “presence of liver disorder”. The experiment results obtained
for various problems by using leave-one-out cross-validation are presented below
(Table 1). In the table heading, n is the number of attributes, l is the number of
objects (the size of the training sample),err is the classification error rate and lc
is the number of classified objects ( l − lc = the number of failed classifications).
Problem n l lc err
1. Liver Disorders 6 345 20 15.00%
2. Glass identification 9 146 25 20.00%
3. Wine 13 130 47 08.50%
4. Wine quality 11 310 51 09.80%
5. Two norm 20 354 109 07.30%
6. Lung cancer 8 137 9 11.10%
7. Cirrhosis 19 276 29 34.48%
8. Cloud-seeding 5 108 6 50.00%
Table 1. Experimental results
The algorithm was updated with the following modifications in definitions of
hypothesis and classification:
1. Hypothesis modification: attributes observed for ”almost” all objects of the
particular class were added to the hypothesis. It was ensured that the ratio
of objects which did not comply with the hypothesis in the same class did
not exceed the value P (a new algorithm parameter) and obviously there was
no guarantee that the hypothesis is not contained in descriptions of objects
of the opposite class.
2. Introduction of an inter-object metric and modification of the classification
procedure: the ”distance” between objects increases as they reveal difference
in a larger number of coordinates. We compute the distance of the object
being classified by positive and negative hypotheses and normalize it by
the number of attributes (or 1s in binary represenation) in each hypothesis.
The object is classified to the nearest class in contexts of the metric defined
above.
3. Attribute weighting: an attribute is assigned a weight which increases with
the number of 1s in the corresponding column.
The modified algorithm was applied to the considered problems. The exper-
imental results for P = 0.2 are given in Table 2.
4.2 Classification Using Biclustering
Biclustering can be used for classification upon data scaling. For this purpose we
select informative objects which are included in biclusters with density greater
than threshold ρmin . Hypotheses are generated using these objects. This ap-
proach avoids noise effects during learning step [15]. The difference between
100 Olga Prokasheva, Alina Onishchenko, Sergey Gurov
Problem n l lc err
1. Liver Disorders 6 345 79 39.2%
2. Glass identification 9 146 64 25.00%
3. Wine 13 130 87 14.9%
4. Wine quality 11 310 142 17.60%
5. Two norm 20 354 224 15.10%
6. Lung cancer 8 137 36 36.10%
7. Cirrhosis 19 276 136 33.30%
8. Cloud-seeding 5 108 37 43.20%
Table 2. Experimental results for the modified algorithm
proposed algorithm and simple FCA algorithm resides only in the second step:
hypotheses are now generated using only informative objects selected by biclus-
tering. The method has two adjustable parameters: the bicluster density ρmin
and the ratio P of objects which do not satisfy classical hypotheses. The parame-
ter ρmin affects the generation of hypotheses. If its value is too small, hypothesis
generation is tainted by noisy attributes and outliers. If its value is too large,
the hypothesis will have to meet excessively stringent requirements. It may be
efficient to use a range of values for ρmin and thus focus on the main objects,
skipping the marginal ones. This method has been tested, but it failed to pro-
duce a significant improvement of the classification performance, which will be
later explained by the specific features of the particular problem.
The parameter P affects the ratio of objects which do not satisfy the hy-
potheses of the same class. When the parameter P is close to zero, hypotheses
are generated in accordance with the classical definitions: they include only the
attributes that are observed for all the objects of the given class. The difficulty
is that hypotheses may become ”non-representative” for the given class. If the
parameter P is taken too large, the hypotheses will require that the control ob-
ject has a large number of attributes, which again may impose an excessively
stringent requirement on hypotheses. In a certain sense, this is the well-known
overfitting effect often observed in pattern recognition.
The experimental results assessed by leave-one-out cross-validation are pre-
sented in Table 3 and Table 4. In the table heading, n is the number of attributes,
l is the number of objects (the size of the training sample). The columns present
the solution results obtained with the algorithm parameters (the threshold ρ
and the proportion P ) optimized by two criteria: the classification error rate
err (Table 3) and the number of classified objects lc ( l − lc = the number of
failed classifications) (Table 4). The local optimization of the algorithm param-
eters was carried out by the GaussSeidel method, their optimal values ρmin and
P ∗ are shown together with err.
According to the experimental results, the lower is the error rate, the smaller
is the number of classified objects. We can construct an algorithm with zero
error rate, but the ratio of classified objects will be also small. We apply such
an algorithm for all considered objects. The results are shown in Table 5, where
Classification Methods Based on Formal Concept Analysis 101
Problem n l lc err ρmin P ∗
1. Liver Disorders 6 345 22 13.6% 0.30 0.01
2. Glass identification 9 146 28 10.00% 0.15 0.05
3. Wine 13 130 76 02.00% 0.25 0.05
4. Wine quality 11 310 83 08.40% 0.25 0.05
5. Two norm 20 354 206 12.10% 0.15 0.15
6. Lung cancer 8 137 18 05.50% 0.01 0.01
7. Cirrhosis 19 276 33 21.00% 0.05 0.05
8. Cloud-seeding 5 108 7 28.00% 0.15 0.05
Table 3. Experimental results. Classification error rate is optimized.
Problem n l lc err ρmin P ∗
1. Liver Disorders 6 345 79 29.1% 0.30 0.20
2. Glass identification 9 146 59 16.90% 0.30 0.20
3. Wine 13 130 85 08.20% 0.30 0.20
4. Wine quality 11 310 141 13.50% 0.30 0.20
5. Two norm 20 354 233 15.20% 0.30 0.20
6. Lung cancer 8 137 98 25.50% 0.05 0.05
7. Cirrhosis 19 276 83 37.79% 0.30 0.20
8. Cloud-seeding 5 108 20 30.00% 0.15 0.15
Table 4. Experimental results. The number of classified objects is optimized.
lc is the number of classified objects (the ratio is in brackets). The efficiency of
FCA-based algorithm was compared with that of classical SVM-algorithm [16].
Each dataset was divided into training sample (80% of objects) and test sample
(20% of objects). Table 5 SV M err shows the error rate of the SVM-algorithm,
SV M err on lc is the error rate on objects which were classified by the rigor-
ous FCA-based method. Zero error rate was attained with classical hypotheses
Problem l lc ρmin P ∗ SV M err SV M err on lc
1. Liver Disorders 345 18 (5.2%) 0.15 0 34.78% 22.2%
2. Glass identification 146 22 (15%) 0.15 0 31.03% 4.55%
3. Wine 130 45 (35%) 0.25 0 7.69% 2.22%
4. Wine quality 130 49 (5.8%) 0.1 0 35.48% 6.12%
5. Two norm 354 103 (29%) 0.03 0 3.85% 0%
6. Lung cancer 137 9 (6.50%) 0.01 0 40.74% 0%
7. Cirrhosis 276 24 (9.00%) 0.05 0 9.8% 12.5%
8. Cloud-seeding 108 5 (4.6%) 0.15 0 40.91% 25.00%
Table 5. Experimental results with zero error rate.
(P =0) from objects with low density (ρmin ≤ 0.25). This rigorous algorithm
can be applied to problems with high error costs. It is more likely to refuse
classification than make wrong decisions.
102 Olga Prokasheva, Alina Onishchenko, Sergey Gurov
5 Conclusions
FCA provides a convenient tool for formalizing symbolic machine learning and
classification models. We studied hypothesis-based classification in different ar-
eas without special modifications for each dataset, using a simple binarization
(scaling) of numerical data. Our results suggest the following conclusions:
1. Application of biclustering with parameter optimization made a very slight
improvement in the quality of classification compared to the updated FCA
algorithm (only by 3% in problem 1).
2. In all cases there was an unacceptably high rate of classification failures.
3. In all cases there was an unacceptably high error rate.
4. Attempts to fine-tune the algorithm parameters with the objective of reduc-
ing the failure rate were generally accompanied by increasing in the number
of errors, although in some cases (problem 8) the error rate increased only
slightly; the number of classifiable objects in these cases increased substan-
tially (problem 6).
5. The classical FCA-based algorithm can produce accurate classification, but
it refuses to classify the majority of test sample.
The analysis of the hypotheses generated with various parameter values and
different optimization criteria has shown that hypotheses of different classes are
often included in one another. We can naturally assume that if the classes show
less tendency to diffuse into one another, biclustering and the classical FCA
method would produce more impressive results. The relative location of classes
is improved in pattern recognition theory by methods that involve transforma-
tion of the attribute space. In these cases, data compactification methods may
be effectively applied [14]. We can reasonably assume that another scaling algo-
rithms with floating-size intervals and interval-length optimization may improve
classification results compared to those we have obtained with the simplest scal-
ing. Our analysis of FCA-based classification provides the following conclusions:
1. For the chosen universal scaling procedure the classification results are far
from being optimal. Individual scaling for each problem may improve clas-
sification quality.
2. FCA-based classification methods without modification and/or thorough
preprocessing of data are usable only for preliminary classification.
3. A well-known idea for the modification of the direct FCA approach is to
develop hypothesis generation methods. It is useful to allow for the specific
features of the particular subject area and to fine-tune hypotheses and the
algorithms by using e.g. parameters ρmin , P*, σmin .
4. It is also possible to develop and apply sharper classification rules, e.g. by
weighting objects, attributes, hypotheses, etc.
5. A promising approach is to use FCA-based methods to transform the at-
tribute space, in particular using data compactness estimates.
6. concept-based methods are appropriate for classification problems with high
error costs, e.g. in medical, security and military applications.
Classification Methods Based on Formal Concept Analysis 103
An important step in many data classification problems is the selection of a
suitable similarity measure. We decided to investigate how different similarity
metric described in [20, 23] affects the quality of hypothesis-based classification
method. The majority of pattern recognition methods use the metric information
about objects: methods based on distances, potential functions, dividing surface,
the algebraic approach, etc. In these methods the amount of information about
classes either is fairly used at all. The strength of FCA-based classification prob-
lems is in the identification and use of these particular data, but the metric
information about the feature space is lost. Thus, in pattern recognition the
classic discriminant methods and method based on FCA are at opposite poles
w.r.t. metrics. In the FCA the metric information appears in a weak form as the
result of scaling, which accounts for “distances” between attributes. It seems
that the success in the development of FCA-based recognition methods will be
related to the introduction of information about metric properties of feature
spaces. Our future work will be focused on developing and applying sharper
classification rules with modifications of similarity measure.
References
1. B. Ganter, R. Wille: Formal Concept Analysis: Mathematical Foundations. Springer,
Berlin/Heidelberg (1999).
2. S.O. Kuznetsov: Mathematical aspects of concept analysis. In: Journal of Mathe-
matical Science, Vol. 80, Issue 2, pp. 1654–1698, (1996).
3. S.O. Kuznetsov: Complexity of Learning in Concept Lattices from Positive and
Negative Examples. Discrete Applied Mathematics, 2004, No. 142(1–3), pp. 111-
125.
4. G. Birkhoff: Lattice Theory. AMS, Providence, 3rd edition (1967).
5. O. Ore: Theory of Graphs. American Mathematical Society, Providence (1962).
6. S.I. Gurov: Ordered Sets and Universal Algebra [in Russian], MGU, Moscow (2004).
7. V.K. Finn: The Synthesis of Cognitive Procedures and the Problem of Induction.
Autom. Doc. Math. Linguist., 43, 149-195 (2009).
8. V.K. Finn: On machine-oriented formalization of plausible reasoning in the style of
F. Bacon and D.S. Mill [in Russian], Semiotika i Informatika, 20, 35–101 (1983).
9. S.C. Madeira and A.L. Oliveira: Biclustering Algorithms for Biological Data Anal-
ysis: A Survey. IEEE/ACM Trans. Comput. Biol. Bioinformatics 1, 24–45 (2004).
10. B.G. Mirkin: Mathematical Classification and Clustering. Kluwer (1996).
11. D.I. Ignatov, S.O. Kuznetsov: Frequent Itemset Mining for Clustering Near Du-
plicate Web Documents. In: Proc. 17th Int. Conf. on Conceptual Structures (ICCS
2009), LNAI (Springer), Vol. 5662, 185–200, 2009.
12. D.I. Ignatov, S.O. Kuznetsov, R.A. Magizov and L.E. Zhukov: From Triconcepts
to Triclusters. In: Proc. 13th Int. Conf. on Rough Sets, Fuzzy Sets, Data Mining,
and Granular Computing (RSFDGrC 2011), LNCS (Springer), Vol. 6743, 257–264,
2011.
13. S.O. Kuznetsov: Stability as an Estimate of the Degree of Substantiation of Hy-
potheses on the Basis of Operational Similarity. In: Nauchno-Tekhnicheskaya Infor-
matsiya, Seriya 2, Vol. 24, No. 12, pp. 21–29, 1990.
14. S.I. Gurov, N.S. Dolotova, I.N. Fatkhutdinov: Noncompact recognition problems.
Circuit design according to E. Gilbert. In: Spectral and Evolution Problems: Proc.
104 Olga Prokasheva, Alina Onishchenko, Sergey Gurov
17 th Crimean Autumn Mathematical School-Symposium, Simferopol, Crimean Sci-
entific Center of Ukrainian Academy of Sciences, 17, pp. 37–44 (2007).
15. A.A. Onishchenko, S.I. Gurov: Classification based on Formal Concept Analysis
and Biclustering: possibilities of the approach. In: Computational Mathematics and
Modeling, Vol. 23, No. 3, July, pp. 329–336 (2012).
16. C. Cortes, V. Vapnik: Support–vector networks. In: Machine Learning, September
1995, Volume 20, Issue 3, pp. 273–297.
17. N. Meddouri, M. Meddouri: Classification Methods based on Formal Concept Anal-
ysis, — CLA 2008 (Posters), pp. 9–16, Palacky University, Olomouc, (2008).
18. S. I. Gurov: Boolean Algebras, Ordered Sets, Lattices: Definitions, Properties,
Examples [in Russian], KRASAND, Moscow (2012).
19. S.O. Kuznetsov: Machine learning on the basis of formal concept analysis Automa-
tion and Remote Control 62 (10),pp. 1543–1564.
20. F. Alqadah, R. Bhatnagar: Similarity measures in formal concept analysis, Annals
of Mathematics and Artificial Intelligence, 61(3), 245–256 (2011).
21. S.O. Kuznetsov: On Stability of a Formal Concept. Annals of Mathematics and
Artificial Intelligence, Vol. 49, pp.101–115, 2007.
22. L. Breiman, J.H. Friedman, R.A. Olsen, and C.J. Stone: Classification and Regres-
sion Trees, Wadsworth Int. Group, Belmont, CA (1984).
23. F. Dau, J. Ducrou, P.W. Eklund: Concept Similarity and Related Categories in
SearchSleuth. ICCS 2008: pp. 255–268
24. M. Sahami: Learning classification Rules Using Lattices. N. Lavrac and S. Wrobel
eds., pp. 343–346, Proc ECML, Heraclion, Crete, Greece (Avril 1995).
25. C. Caprineto, G. Romano: GALOIS An order-theoretic approach to conceptual
clustering. In proceedings of ICML93, pp. 33–40, Amherst, USA (July 1993).
26. P. Njiwoua, E.M. Nguifo: Forwarding the choice of bias LEGAL-F Using Feature
Selection to Reduce the complexity of LEGAL. In Proceedings of BENELEARN-97,
ILK and INFOLAB, Tilburg University, the Netherlands, pp. 89–98, 1997.
27. Zhipeng Xie, Wynne Hsu, Zongtian Liu, Mong Li Lee: Concept Lattice based
Composite Classifiers for high Predictability. Artificial Intelligence, vol. 139, pp.
253–267, Wollongong, Australia (2002).
28. S. Guillas, K. Bertet, J-M. Ogier: Extension of Bordats algorithm for attributes.
Concept Lattices and Their Applications: CLA07, Montpellier, France (2007).
29. B. Douar, C. Latiri, Y. Slimani: Approche hybride de classification supervisee
base de treillis de Galois: application la reconnaissance de visages. In: Extraction
et Gestion des Connaissances (EGC08), 309–320, Nice, France (2008).
30. P. Njiwoua, Mephu Nguifo E.: Ameliorer lapprentissage partir dinstances grace
linduction de concepts: le systeme CIBLe, Revue dIntelligence Artificielle (RIA),
vol. 13, 2, 1999, pp. 413–440, Hermes Science.
31. E.M. Nguifo, P. Njiwoua: Treillis de concepts et classification supervise. Technique
et Science Informatiques: TSI, Volume 24, Issue 4, pp. 449–488 (2005).
32. M. Maddouri: Towards a machine learning approach based on incremental concept
formation. Intelligent Data Analysis, Volume 8, Issue 3, pp. 267–280 (2004).