1 Introduction
2 Relational data mining and granular computing
2.1 Relational data mining
2.1.1 Graph-based relational data mining approaches
2.1.2 Formal concept analysis for relational data mining
2.2 Granular computing
2.3 Granular computing in relational data mining
3 Constrained sums of information systems
3.1 Relational data representation
3.2 Relational information granule construction
3.3 Relational pattern construction
3.4 Illustrative example
Id | Age | Gender | Income | Class |
---|---|---|---|---|
Customer | ||||
1 | 30 | Male | 1500 | Yes |
2 | 33 | Female | 2500 | Yes |
3 | 30 | Female | 1800 | No |
4 | 30 | Female | 1800 | Yes |
5 | 26 | Female | 2500 | Yes |
6 | 29 | Male | 3000 | Yes |
7 | 30 | Male | 1800 | No |
Id | Name | Price | ||
---|---|---|---|---|
Product | ||||
1 | Bread | 2.00 | ||
2 | Butter | 3.50 | ||
3 | Milk | 2.50 | ||
4 | Tea | 5.00 | ||
5 | Coffee | 6.00 | ||
6 | Cigarettes | 6.50 |
Id | Cust_id | Prod_id | Amount | Date |
---|---|---|---|---|
Purchase | ||||
1 | 1 | 1 | 1 | 24/06 |
2 | 1 | 3 | 2 | 24/06 |
3 | 2 | 1 | 1 | 25/06 |
4 | 2 | 3 | 1 | 26/06 |
5 | 4 | 6 | 1 | 26/06 |
6 | 4 | 2 | 3 | 27/06 |
7 | 5 | 5 | 2 | 27/06 |
8 | 6 | 4 | 1 | 27/06 |
4 Granular association rule approach
4.1 Relational data representation
4.2 Relational information granule construction
4.3 Relational pattern construction
-
The support of GR is \(supp_c(GR)=ssupp(GR)\).
-
The confidence of GR is \(conf_{c}(GR)=conf(GR)=1\).
-
The support of GR is \(supp_{lp}(GR)=\frac{|\{x\in e(g_1):e(g_2)\subseteq R(x)\}|}{|U|}\).
-
The confidence of GR is \(conf_{lp}(GR)=\frac{|\{x\in e(G_1):e(g_2)\subseteq R(x)\}|}{|e(g_1)|}\).
-
The support of GR is \(supp_{rp}(GR)=ssupp(GR)\).
-
The confidence of GR is\(conf_{lp}(GR)=min\{\frac{|e(g_2)\cap R(x)|}{e(g_2)}:x\in e(g_1)\}\).
-
The support of GR with respect to a target confidence threshold \(tc\in (0,1]\) is \(supp(GR,tc)=\frac{|\{x\in e(g_1):\frac{|R(x)\cap e(g_2)|}{|e(g_2)|}\ge tc\}|}{|U|}\).
-
The source confidence of GR with respect to tc is \(sconf(GR,tc)=\frac{|\{x\in e(g_1):\frac{|R(x)\cap e(g_2)|}{|e(g_2)|}\ge tc\}|}{|e(g_1)|}\).
4.4 Illustrative example
5 Generalized related set based approach
5.1 Relational data representation
-
\(D_T\) and \(D_B\) denote, respectively, the sets of target and background relations of database D (i.e. a set of all relations);
-
\(U_{D_T}=\bigcup _{R\in D_T} R\) and \(U_{D_B}=\bigcup _{R\in D_B} R\) be, respectively, the set of all target and background objects of database D
-
\(A_{D_T}=\bigcup _{R\in D_T} A_R\)7 and \(A_{D_B}=\bigcup _{R\in D_B} A_R\) be, respectively, the set of all attributes of the target and background relations of database D.
5.2 Relational information granule construction
5.3 Relational pattern construction
5.4 Illustrative example
6 Description language based approach
6.1 Relational data representation
6.2 Relational information granule construction
6.3 Relational pattern construction
6.4 Illustrative example
7 Discussion
7.1 Comparative study
Framework | |
---|---|
Data type
| |
F1 | Propositional database; relational database where join tables links two tables only |
F2 | Many-to-many case where the join table consists of two foreign keys only |
F3 | Relational database |
F4 | Propositional database, relational database |
Language
| |
F1 | Attribute-value language |
F2 | Attribute-value language |
F3 | Relational language |
F4 | Extended attribute-value language |
Task
| |
F1 | Hierarchical modeling of complex pattern |
F2 | Association discovery |
F3 | Multi-task (association discovery, classification, clustering) |
F4 | Multi-task (association discovery, classification, clustering) |
Application
| |
F1 | Multi-agent system based problem solving, e.g. failure diagnosis of the space robotic arm |
F2 | Recommender system, e.g. cold-start system |
F3 | General-purpose, e.g. building a decision support system |
F4 | General-purpose, e.g. building a decision support system |
7.2 Towards building a granular computing based system form mining relational data
-
Non-relational\(\rightarrow \)relational Upgrading a data mining algorithm to a relational case (Van Laer and De Raedt 2001).
-
Non-relational\(\rightarrow \)granular Constructing a granule description language (Skowron and Stepaniuk 2001).
-
relational\(\rightarrow \)granular-relational Constructing a relational granule description language (Hońko 2015a).
-
Granular\(\rightarrow \)granular-relational Upgrading a granular data mining framework to a relational case (Hońko 2014).