J. Han, and Y. Fu, Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases, in Proceedings AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), Seattle, WA, 157-168. July, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....in and resets the SrcIP attribute of the remaining 26 alarms to its original value, namely ip1. Finally, the DstIP attribute is generalized twice, the alarm (ip1,External IPs, 26) is reported, and processing ends. Our modified version of AOI also supports dynamic generalization hierarchies [22]. Dynamic generalization hierarchies are constructed at run time to fit the data distribution. For example, instead of a static generalization hierarchy that imposes concepts such as morning , evening , night , etc. we dynamically construct generalization hierarchies for the time attribute. ....

J. Han and Y. Fu. Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases. In Workshop on Knowledge Discovery in Databases, pages 157--168, 1994.

....In contrast, discretization is aimed to reducing the number of distinct values for a given attribute, particularly for analysis methods requiring discrete attribute values. Possible solutions for discretization are # histogram based discretization, # discretization based on concept hierarchies [13], # entropy based discretization [9] Considering only the histogram based approach numeric values could be replaced by a representative discrete value associated with the containing bucket as already discussed in Section 4.3. Here, both kinds of histograms are applicable. An alternative ....

J. Han and Y. Fu. Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases. In AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), Seattle, WA, pages 157--168, 1994.

....run time. Another important feature of our approach is that we consider multiple concept hierarchies at the same time. Such hierarchies are either provided by the administrator (thus representing some kind of domain knowledge) or can be discovered using data using clustering techniques (see, e.g. [Eve73, HF94]) Finally, and most importantly, we introduce the notion of interestingness measure to determine the relevance of feature sets in the discovery process. This measurement can be specified by the administrator depending on the type and granularity of policy s he is interested in. 1.1 Related Work ....

....above, even these approaches turn out to generate too many fine grained profiles and policies to be useful in complex information system scenarios. Multiple concept hierarchies have been introduced in the data mining domain for deriving typical patterns of data at different levels of abstraction [CCH91, HF94, HF95, SA95]. In the approach described in this paper, we extend the usage of multiple concept hierarchies in several ways. First, we allow different types of concepts in a single hierarchy, thus allowing administrators to embed knowledge in a more natural and intuitive way. Second, our framework is more ....

J Han and Y Fu. Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases. AAAI'94 Workshop on Knowledge Discovery in Databases, pages 157--168, July 1994.

....Reduction by Concept Ascension. The last two methods are relatively simply understood and accommodated but suffer from the problem that the information capacity reduces rapidly as attributes and tuples are removed. The idea of accommodating hierarchical domains has been discussed elsewhere [3 5, 10] and provides a mechanism whereby the information capacity of a summary dataset may degrade more slowly for a similar reduction in space. Briefly, the idea is to provide, commonly through user supplied hierarchies although they may also be generated a priori by autonomous procedures, higher level ....

Han, J. and Fu, Y. Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases, 157-168. 1994.

....term, leading to unification failure if an object is postulated to belong to both sorts. This opens a whole area of research for generalizing our spanning set framework for encoding extended partial orders. Data Mining. Tree shaped conceptual hierarchies have been proposed for use in data mining [13, 81, 82]. There exists a great potential for generalizing these techniques to use partial orders, and even extended partial orders. Reference Constraints. To fully demonstrate the utility of individual level inheritance, reference constraints must be implemented in a logic programming system. ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, 1994.

....hierarchy Figure 4.1: The Concept Hierarchies of (a) REGION and (b) CATEGORY is formed by grouping the continent(level 0) and the country(level 1) columns and the category hierarchy is formed by grouping the product line(level 0) and product type(level 1) columns. Previous works [Fis87, CC94, HF94] have described how to automatically generate concept hierarchies for numerical columns in the database table based on data semantics and or data distribution statistics. Thus, a numerical column revenue in the database table can actually form a concept hierarchy as illustrated in the following ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

....hierarchies or adjust some existing hierarchies for certain tasks. The methods for automatic generation of concept hierarchies for numerical attributes based on data distributions and for dynamic refinement of a given or generated concept hierarchy based on a learning request are introduced in [22, 18]. Other interesting studies on automatic generation of hierarchies for categorical data can be found in [15, 17, 29, 32, 33] 2.2 Pattern Discovery There are lots of works done in the area of Artificial Intelligence related to pattern discovery in sequences of events. The problem considered in ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

....in a very straightforward manner as in the case of geographical data. If neither is feasible, methods are available to create concept hierarchies from data automatically and to modify existing hierarchies to suit the current mining task if for example only a part of the data is being examined [18]. This is necessary when the current structure is too general, specific or unbalanced and thus causes distorted results. Concept hierarchies have been used in classification mining before, the most prominent example of which is attribute oriented induction that is realized in the DBLEARN system ....

Jiawei Han and Yongjian Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), Seattle, WA, July 1994.

....Moreover, the generalized relation can be further analyzed by integration with other machine learning methods [7] including ID 3 [19] Cluster 2 [16] etc. The system also performs automatic generation of conceptual hierarchies for numerical attributes and dynamic conceptual hierarchy adjustment [5] for all the attributes based on the statistical distribution of the set of relevant data, which produces desirable generalized results. DBMiner offers both graphical and SQL like interfaces [7] For example, to characterize Computer Science grants in the NSERC9J database in relevance to ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157 168, Seattle, WA, July 1994.

....on user s learning requirements, data semantics and or data distribution statistics. Moreover, a given concept hierarchy may not be best suited for a particular learning task, which therefore often needs to be dynamically refined based on data distribution statistics for desired learning results [26]. For example, if the learning requirement is to analyze the birth place of the students of Simon Fraser University, the level 1 (top level) concepts could be: B.C, other provinces in Canada, foreign . However, if it is to analyze the birth place of the faculty of the university, the appropriate ....

....cube generation. a) Compute the desired level Li for each dimension a based on its dimension threshold T (automatic hierarchy generation can be performed for numerical data if there is none, and dynamic hierarchy adjustment can be performed for both numerical and nonnumerical data, if desired [26]) Notice that ai should be removed if there is a large set of distinct values in ai in 1420 but there is no generalization operator on (b) Determine the mapping pairs (v, v t) for each dimension a, where v is a distinct value of ai, and v t is its corresponding generalized value at level Li. 4. ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'9 Workshop on Knowledge Discovery in Databases (KDD'9), pages 157-168, Seattle, WA, July 1994.

....be done with higher accuracy. ChiMerge is designed solely for classification in which several classification attributes must be pre specified. Otherwise, the 2 value is impossible to be obtained CHAPTER 2. RELATED WORK 12 if there is no any classification attributes given. In 1994, Han and Fu[25] reported a study on the automatic generation and dynamic adjustment of concept hierarchies based on data mining tasks. The role of concept hierarchies in the attribute oriented induction is clarified and several algorithms are developed for the generation and adjustment of concept hierarchies. ....

....hierarchies is described. An algorithm based on hierarchical clustering with order constraint is proposed in x4.2.2, and another algorithm based on partitioning clustering is developed in x4.2.3. Performance analysis and quality comparison are presented in x4.2.4. 4.2. 1 Basic Algorithm Han and Fu[25] reported an algorithm for the automatic generation of numerical hierarchies. The idea is based on the consideration that it is desirable to present rules or regularities by a set of nodes with relatively even data distribution, i.e. not a blend of very big nodes and very small nodes at the same ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases(KDD'94), Seattle, WA, 157-168, 1994.

....induction which indicates the number of original records covered by a generalized record. Using the example query given in Section 2.1. 3, we will take mining characteristic rules as an example to illustrate the attribute oriented induction which is performed in the steps described below [33]: CHAPTER 2. RELATED WORK 14 Name Gender Age Birth place Department GPA C. Smith female 18 Montreal Computer Science 3.92 M. Jordan male 20 Chicago Engineering 2.36 G. Tong male 26 Beijing Math 3.25 Delta Delta Delta Delta Delta Delta Delta Delta Delta Delta Delta Delta Delta ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994. BIBLIOGRAPHY 118

....and be stored in the form of relations in the same database. Moreover, they can be adjusted dynamically based on the distribution of the set of data relevant to the data mining task. Also, hierarchies for numerical attributes can be constructed automatically based on data distribution analysis [7]. 3 DMQL and Interactive Data Mining DBMiner offers both an SQL like data mining query language, DMQL, and a graphical user interface for interactive mining of multiple level knowledge. Example 1. To characterize CS grants in the NSERC96 database related to discipline code and amount category ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

....levels. The information about concept hierarchies can be provided by knowledge engineers or domain experts or be discovered automatically or semi automatically using knowledge discovery tools based on the statistics of data distribution in databases and the relationships among different attributes [10]. Many concept hierarchies are implicitly stored in the database. For example, the hierarchical relationship among city , province and country attributes are usually stored in the database and can be made explicit at the schema level by indicating a part of hierarchy: city ae province ae ....

....constructed dynamically without prior knowledge based on the value range distribution in the database. For other hierarchies found in the KRDB, modification can be performed dynamically based on the statistics of current relevant data sets and user preference in order to extract interesting rules [10]. For example, to extract the interesting relationships between GPA and Birth place, a given hierarchy can be modified dynamically to allow more detailed distributions of Birth places in nearby provinces or countries than remote ones. A prime relation maintains the relationships among generalized ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

.... study, we assume that multiple levels of concepts are organized in the form of concept hierarchies which are provided in the system for mining rules at multiple concept levels, however, the concept hierarchies can also be dynamically adjusted and or automatically generated for flexible data mining [5]. To confine our study, we assume the rules to be discovered are conjunctive rules, i.e. a set of conjuncts in both the rule head and body. Moreover, the predicate variable in the meta rules can only be instantiated against database schema (attributes) Furthermore, each predicate variable in a ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

....(i.e. meta rules) to indicate meaningful or desired mappings, such as fcontent spec, brand, categoryg ae fcontent spec, categoryg ae category , etc. Concept hierarchies may not exist for numerical valued attributes but can be automatically generated according to data distribution statistics [9, 6]. For example, a hierarchy for the price range of sales items can be generated based on the distribution of price values. Moreover, a given concept hierarchy for numerical or nonnumerical data can be dynamically adjusted based on data distribution [9] For example, if there are many distinct ....

....according to data distribution statistics [9, 6] For example, a hierarchy for the price range of sales items can be generated based on the distribution of price values. Moreover, a given concept hierarchy for numerical or nonnumerical data can be dynamically adjusted based on data distribution [9]. For example, if there are many distinct country names in the attribute place made , countries can be grouped into continents, such as Asia, Europe, South America, etc. Moreover, if most fresh food products are from B.C. and Northwest America, the geographic hierarchy should be adjusted to ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

....later in the classification process, and enables the handling of noisy and exceptional data. Concept hierarchies may be provided by domain experts or database administrators, or may be defined using the database schema [11] Concept hierarchies for numeric attributes can be generated automatically [12]. In addition to allowing the substantial reduction in size of the training set, concept hierarchies allow the representation of data in the user s vocabulary. Hence, aside from increasing efficiency, attribute oriented induction may result in classification trees that are more understandable, ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, 1994.

....it can be defined on a single attribute or on a set of related attributes, and it can be in the shape of a balanced tree, a lattice or a general DAG. Furthermore, a given concept hierarchy can be adjusted dynamically based on the analysis of the statistical distribution of the relevant data sets [13]. 3.2 Generalization of Simple Numerical Values Generalization of numerical attributes can be performed similarly but in a more automatic way by the examination of data distribution characteristics [1, 9, 5] In many cases, it may not require any predefined concept hierarchies. For example, the ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

....land use concept hierarchy form of concept hierarchies. In the case of spatial databases, there can be two kinds of concept hierarchies, non spatial and spatial. Concept hierarchies can be explicitly given by the experts, or in some cases they can be generated automatically by data analysis [26]. An example of a concept hierarchy for agricultural land use is shown in Figure 2. As we ascend the concept tree, information becomes more and more general, but still remains consistent with the lower concept levels. For example, in Figure 2 both jasmine and basmati can be generalized to the ....

J. Han and Y. Fu. Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pp. 157--168, Seattle, WA, July 1994.

....schema level (i.e. meta rules) to indicate meaningful or desired mappings, such as fcontent, brand, categoryg ae fcontent, categoryg ae category , etc. Concept hierarchies may not exist for numerical valued attributes but can be automatically generated according to data distribution statistics [8, 5]. For example, a hierarchy for the price range of sales items can be generated based on the distribution of price values. Moreover, a given concept hierarchy for numerical or nonnumerical data can be dynamically adjusted based on data distribution [8] For example, if there are many distinct ....

....according to data distribution statistics [8, 5] For example, a hierarchy for the price range of sales items can be generated based on the distribution of price values. Moreover, a given concept hierarchy for numerical or nonnumerical data can be dynamically adjusted based on data distribution [8]. For example, if there are many distinct country names in the attribute place made , countries can be grouped into continents, such as Asia, Europe, South America, etc. Moreover, if most fresh food products are from B.C. and Northwest America, the geographic hierarchy can be automatically ....

[Article contains additional citation context not shown here]

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In AAAI'94 Workshop on Knowledge Discovery in Databases, pp. 157--168, Seattle, WA, July 1994.

....and be stored in the form of relations in the same database. Moreover, they can be adjusted dynamically based on the distribution of the set of data relevant to the data mining task. Also, hierarchies for numerical attributes can be constructed automatically based on data distribution analysis (Han Fu 1994). DMQL and Interactive Data Mining DBMiner offers both an SQL like data mining query language, DMQL, and a graphical user interface for interactive mining of multiple level knowledge. Example 1. To characterize CS grants in the NSERC96 database related to discipline code and amount category ....

Han, J., and Fu, Y. 1994. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), 157--168.

....on user s learning requirements, data semantics and or data distribution statistics. Moreover, a given concept hierarchy may not be best suited for a particular learning task, which therefore often needs to be dynamically refined based on data distribution statistics for desired learning results [26]. For example, if the learning requirement is to analyze the birth place of the students of Simon Fraser University, the level 1 (top level) concepts could be: fB.C, other provinces in Canada, foreigng. However, if it is to analyze the birth place of the faculty of the university, the appropriate ....

....generation. a) Compute the desired level L i for each dimension a i based on its dimension threshold T i (automatic hierarchy generation can be performed for numerical data if there is none, and dynamic hierarchy adjustment can be performed for both numerical and nonnumerical data, if desired [26]) Notice that a i should be removed if there is a large set of distinct values in a i in W 0 but there is no generalization operator on a i . b) Determine the mapping pairs hv; v 0 i for each dimension a i , where v is a distinct value of a i , and v 0 is its corresponding generalized value ....

J. Han and Y. Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pages 157--168, Seattle, WA, July 1994.

No context found.

J. Han, and Y. Fu, Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases, in Proceedings AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), Seattle, WA, 157-168. July, 1994.

No context found.

Jiawei Han and Yongjian Fu. Dynamic generation and refinement of concept hierarchies for knowledge discovery in databases. In AAAr94 Workshop on Knowledge Discovery in Databases (K.DD'94), Seattle, WA, July 1994.

No context found.

Han, J. and Y. Fu. Dynamic Generation and Refinement of Concept Hierarchies for Knowledge Discovery in Databases In Proc. AAAI'94 Workshop on Knowledge Discovery in Databases (KDD'94), pp. 157-168, Seattle, WA, July 1994.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC