Exploiting the Power of Group Differences: Using Patterns to Solve Data Analysis Problems

Abstract

The series publishes 50-to 150-page publications on topics pertaining to data mining, web mining, text mining, and knowledge discovery, including tutorials and case studies. Potential topics include: data mining algorithms, innovative data mining applications, data mining systems, mining text, web and semi-structured data, high performance and parallel/distributed data mining, data mining standards, data mining and knowledge discovery framework and process, data mining foundations, mining data streams and sensor data, mining multi-media data, mining social networks and graph data, mining spatial and temporal data, pre-processing and post-processing in data mining, robust and scalable statistical methods, security, privacy, and adversarial data mining, visual data mining, visual analytics, and data visualization.

A Review of Supervised Classification based on Contrast Patterns: Applications, Trends, and Challenges

“…A supervised classifier based on contrast patterns uses a collection of contrast patterns to create a model that classifies a query object in a predefined class [47,200]. A pattern is an expression defined in a certain language that describes a collection of objects [46,47,[123][124][125]127]. Usually, a pattern is represented by a conjunction of relational statements (a.k.a items), each with the form:…”

“…where v j is a value in the domain of feature f i , and # is a relational operator from the set {∈, / ∈, =, =, ≤, >} [47,127,200]. For example, [seizures per day ∈ [5,15]] ∧ [N umber of f atigues > 5] ∧ [H eadaches ≤ 10]∧[cognitive impairment = "Mild"] is a pattern describing a collection of people suffering from brain cancer metastases [142].…”

“…For example, [seizures per day ∈ [5,15]] ∧ [N umber of f atigues > 5] ∧ [H eadaches ≤ 10]∧[cognitive impairment = "Mild"] is a pattern describing a collection of people suffering from brain cancer metastases [142]. Let p be a pattern, and C = {C 1 , C 2 , C 3 , ..., C n } a set of classes such that C 1 ∪C 2 ∪ C 3 ∪...∪C n = U ,C i = C i ; then, support (p, C i ,C i ) is the fraction resulting from dividing the number of objects belonging to C i described by p by the total number of objects belonging to C i [46,47,49,127,200].…”

“…We describe the function support using three parameters to be consistent with the notation that we introduced in Section 2.2.2 for describing the pattern's quality measures. Now, a contrast pattern (CP) for a class C i is a pattern p where the support of p for C i is significantly higher than any support of p for every class other than C i [46,47,49,127,200].…”

“…We also remark that while there are a number of similar literature review and survey articles [1,47,62,73,101], and books [48,158], there are a number of differences as explained below:…”