DSpace community: 軍訓室

Discover Sequential Patterns in Incremental Database

title: Discover Sequential Patterns in Incremental Database abstract: The task of sequential pattern mining is to discover the complete set of sequential patterns in a given sequence database with minimum support threshold. But in practice, minimum support some time is defined afterward, or need to be adjusted to discover information that interest to knowledge workers. In the same time, the problem of discover sequential patterns in a incremental database is an essential issue in real world practice of datamining. This paper discusses the issue of maintaining discovered sequential patterns when some information is appended to a sequence database. Many previous works based on Apriori-like approaches are not capable to do so without re-running previously presented algorithms on the whole updated database. We propose a novel algorithm, called DSPID, which takes full advantage of the information obtained from previous mining results to cut down the cost of finding new sequential patterns in an incremental database.

An Algorithm for Mining Strong Negative Fuzzy Sequential Patterns

title: An Algorithm for Mining Strong Negative Fuzzy Sequential Patterns abstract: Many methods have been proposed for mining fuzzy sequential patterns. However, most of conventional methods only consider the occurrences of fuzzy itemsets in sequences. The fuzzy sequential patterns discovered by these methods are called as positive fuzzy sequential patterns. In practice, the absences of frequent fuzzy itemsets in sequences may imply significant information. We call a fuzzy sequential pattern as a negative fuzzy sequential pattern, if it also expresses the absences of fuzzy itemsets in a sequence. In this paper, we proposed a method for mining negative fuzzy sequential patterns, called NFSPM. In our method, the absences of fuzzy itemsets are also considered. Besides, only sequences with high degree of interestingness can be selected as negative fuzzy sequential patterns. An example was taken to illustrate the process of the algorithm NFSPM. The result showed that our algorithm could prune a lot of redundant candidates, and could extract meaningful fuzzy sequential patterns from a large number of frequent sequences.

A Deflected Grid-based Algorithm for Clustering Analysis

title: A Deflected Grid-based Algorithm for Clustering Analysis abstract: The grid-based clustering algorithm, which partitions the data space into a finite number of cells to form a grid structure and then performs all clustering operations on this obtained grid structure, is an efficient clustering algorithm, but its effect is seriously influenced by the size of the cells. To cluster efficiently and simultaneously, to reduce the influences of the size of the cells, a new grid-based clustering algorithm, called DGD, is proposed in this paper. The main idea of DGD algorithm is to deflect the original grid structure in each dimension of the data space after the clusters generated from this original structure have been obtained. The deflected grid structure can be considered a dynamic adjustment of the size of the original cells, and thus, the clusters generated from this deflected grid structure can be used to revise the originally obtained clusters. The experimental results verify that, indeed, the effect of DGD algorithm is less influenced by the size of the cells than other grid-based ones.

A Fuzzy Statistics based Method for Mining Fuzzy Correlation Rules

title: A Fuzzy Statistics based Method for Mining Fuzzy Correlation Rules abstract: Mining fuzzy association rules is the task of finding the fuzzy itemsets which frequently occur together in large fuzzy dataset, but most proposed methods may identify a fuzzy rule with two fuzzy itemsets as interesting when, in fact, the presence of one fuzzy itemsets in a record does not imply the presence of the other one in the same record. To prevent generating this kind of misleading fuzzy rule, in this paper, we construct a new method for finding relationships between fuzzy itemsets based on fuzzy statistics, and the generated rules are called fuzzy correlation rules. In our method, a fuzzy correlation analysis which can show us the strength and the type of the linear relationship between two fuzzy itemsets is used. By using thus fuzzy statistics analysis, the fuzzy correlation rules with the information about that two fuzzy not only frequently occur together in same records but also are related to each other can be generated.

An Axis-shifted Crossover-Imaged Clustering Algorithm

title: An Axis-shifted Crossover-Imaged Clustering Algorithm abstract: With low computation time, the grid-based clustering algorithms are efficient clustering algorithms, but the size of the predefined grids and the threshold of the significant cells are seriously influenced their effects. In grid-based clustering system, the data space is partitioned into a finite number of cells to form a grid structure and then performs all clustering operations on this obtained grid structure. The ADCC algorithm is the first one to use axis-shifted strategy to reduce the influences of the size of the cells and inherits the advantage with the low time complexity. But it still uses the cell-clustering twice, the Axis-shifted Crossover-Imaged Clustering Algorithm, called ACICA+, is proposed to use cell-clustering only once and still has the same results. The main idea of ACICA+ algorithm is to shift the original axis in each dimension of the data space after the image of significant cells generated from the original grid structure have been obtained. Because the shifted grid structure can be considered a dynamic adjustment of the size of original cells and the threshold of significant cells, the new image generated from this shifted grid structure will be used to revise and replace the originally obtained significant cells. Finally the clusters will be generated from this crossover image. The experimental results verify that, indeed, the effect of ACICA+ algorithm is less influenced by the size of the cells than other grid-based algorithms. Finally, we will verify by experiment that the results of our proposed ACICA+ algorithm outperforms than others.

A Crossover-Imaged Clustering Algorithm with Bottom-up Tree Architecture

title: A Crossover-Imaged Clustering Algorithm with Bottom-up Tree Architecture

An Axis-Shifted Grid-Clustering Algorithm

title: An Axis-Shifted Grid-Clustering Algorithm abstract: These spatial clustering methods can be classified into four categories: partitioning method, hierarchical method, density-based method and grid-based method. The grid-based clustering algorithm, which partitions the data space into a finite number of cells to form a grid structure and then performs all clustering operations to group similar spatial objects into classes on this obtained grid structure, is an efficient clustering algorithm. To cluster efficiently and simultaneously, to reduce the influences of the size and borders of the cells, a new grid-based clustering algorithm, an Axis-Shifted Grid-Clustering algorithm (ASGC), is proposed in this paper. This new clustering method combines a novel density-grid based clustering with axis-shifted partitioning strategy to identify areas of high density in the input data space. The main idea is to shift the original grid structure in each dimension of the data space after the clusters generated from this original structure have been obtained. The shifted grid structure can be considered as a dynamic adjustment of the size of the original cells and reduce the weakness of borders of cells. And thus, the clusters generated from this shifted grid structure can be used to revise the originally obtained clusters. The experimental results verify that, indeed, the effect of this new algorithm is less influenced by the size of cells than other grid-based ones and requires at most a single scan through the data.

An Adaptable Deflect and Conquer Clustering Algorithm

title: An Adaptable Deflect and Conquer Clustering Algorithm abstract: The grid-based clustering algorithm is an efficient clustering algorithm, but the effect of the algorithm is seriously influenced by the size of the predefined grids and the threshold of the significant cells. Thus, in this paper, to reduce the influences of the size of the predefined grids and the threshold of the significant cells, we adopt deflect and conquer techniques to propose a new grid-based clustering algorithm, which is called Adaptable Deflect and Conquer Clustering (ADCC) algorithm. The idea of ADCC is to utilize the predefined grids and predefined threshold to identify the significant cells, by which nearby cells that are also significant can be merged to develop a cluster in the first place. Next, the modified grids which are deflected to half size of the grid are used to identify the significant cells again. Finally, the new generated significant cells and the initial significant cells are merged so as to offset the round-off error and improve the precision of clustering task. And we verify by experiment that the performance of our new grid-based clustering algorithm, ADCC, is good.

最大序列型樣的的快速探勘

title: 最大序列型樣的的快速探勘

Mining Negative Fuzzy Sequential Patterns

title: Mining Negative Fuzzy Sequential Patterns

An Adaptive Crossover-Imaged Clustering Algorithm

title: An Adaptive Crossover-Imaged Clustering Algorithm abstract: The grid-based clustering algorithm is an efficient clustering algorithm, but its effect is seriously influenced by the size of the predefined grids and the threshold of the significant cells. The data space will be partitioned into a finite number of cells to form a grid structure and then performs all clustering operations on this obtained grid structure. To cluster efficiently and simultaneously, to reduce the influences of the size of the cells and inherits the advantage with the low time complexity, an Adaptive Crossover-Imaged Clustering Algorithm, called ACICA, is proposed in this paper. The main idea of ACICA algorithm is to deflect the original grid structure in each dimension of the data space after the image of significant cells generated from the original grid structure have been obtained. Because the deflected grid structure can be considered a dynamic adjustment of the size of original cells and the threshold of significant cells, the new image generated from this deflected grid structure will be used to revise the originally obtained significant cells. Hence, the new image of significant cells is projected on the original grid structure to be the crossover image. Finally the clusters will be generated from this crossover image. The experimental results verify that, indeed, the effect of ACICA algorithm is less influenced by the size of the cells than other grid-based algorithms. Finally, we will verify by experiment that the results of our proposed ACICA algorithm outperforms than others.

A Crossover-Imaged Clustering Algorithm with Bottom-Up Tree Architecture

title: A Crossover-Imaged Clustering Algorithm with Bottom-Up Tree Architecture abstract: The grid-based clustering algorithms are efficient with low computation time, but the size of the predefined grids and the threshold of the significant cells are seriously influenced their effects. The ADCC [1] and ACICA+ [2] are two new grid-based clustering algorithms. The ADCC algorithm uses axis-shifted strategy and cell clustering twice to reduce the influences of the size of the cells and inherits the advantage with the low time complexity. And the ACICA+ uses the crossover image of significant cells and just only one cell clustering. But the extension of original significant cell in one crossover image is not easy to find what else clusters it belongs to. The crossover-imaged clustering algorithm with bottom-up tree architecture, called CIC-BTA, is proposed to use bottom-up tree architecture to have the same results. The main idea of CIC-BTA algorithm is to use the bottom-up tree architecture to link the significant cells to be the pre-clusters and combine pre-clusters into one by using semi-significant cells The final set of clusters is the result.

以 FAT 演算法挖掘頻繁學習序列

title: 以 FAT 演算法挖掘頻繁學習序列

彈道飛彈威脅與防禦

title: 彈道飛彈威脅與防禦

國防科技概論

title: 國防科技概論

空軍的組成與中華民國空軍

title: 空軍的組成與中華民國空軍

Mining Strong Positive and Negative Sequential Patterns

title: Mining Strong Positive and Negative Sequential Patterns abstract: In data mining field, sequential pattern mining can be applied in divers applications such as basket analysis, web access patterns analysis, and quality control in manufactory engineering, etc. Many methods have been proposed for mining sequential patterns. However, conventional methods only consider the occurrences of itemsets in customer sequences. The sequential patterns discovered by these methods are called as positive sequential patterns, i.e., such sequential patterns only represent the occurrences of itemsets. In practice, the absence of a frequent itemset in a sequence may imply significant information. We call a sequential pattern as negative sequential pattern, which also represents the absence of itemsets in a sequence. The two major difficulties in mining sequential patterns, especially negative ones, are that there may be huge number of candidates generated, and most of them are meaningless. In this paper, we proposed a method for mining strong positive and negative sequential patterns, called PNSPM. In our method, the absences of itemsets are also considered. Besides, only sequences with high degree of interestingness will be selected as strong sequential patterns. An example was taken to illustrate the process of PNSPM. The result showed that PNSPM could prune a lot of redundant candidates, and could extract meaningful sequential patterns from a large number of frequent sequences.