Browsing by Author "Hong, TP"

Now showing 1 - 20 of 24

A coverage-based genetic knowledge-integration strategy
(2000-07-01)
Wang, CH
;
Hong, TP
;
Chang, MB
;
Tseng, SS
;
資訊工程學系
In this paper, we propose a coverage-based genetic knowledge-integration approach to effectively integrate multiple rule sets into a centralized knowledge base. The proposed approach consists of two phases: knowledge encoding and knowledge integration. In the knowledge-encoding phase, each rule in the various rule sets that are derived from different sources (such as expert knowledge or existing knowledge bases) is first translated and encoded as a fixed-length bit string. The bit strings combined together thus form an initial knowledge population. In the knowledge-integration phase, a genetic algorithm applies genetic operations and credit assignment at each rule-string to generate an optimal or nearly optimal rule set. Experiments on diagnosing brain tumors were made to compare the accuracy of a rule set generated by the proposed approach with that of the initial rule sets derived from different groups of experts or induced by various machine learning techniques. Results show that the rule set derived by the proposed approach is more accurate than each initial rule set on its own. (C) 2000 Elsevier Science Ltd. All rights reserved.
A fuzzy inductive learning strategy for modular rules
(1999-04-01)
Wang, CH
;
Liu, JF
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
In real applications, data provided to a learning system usually contain linguistic information which greatly influences concept descriptions derived by conventional inductive learning methods. The design of learning methods to learn concept descriptions in working with vague data is thus very important. In this paper, we apply fuzzy set concepts to machine learning to solve this problem. A fuzzy learning algorithm based on the maximum information gain is proposed to manage linguistic information. The proposed learning algorithm generates fuzzy rules from "soft" instances, which differ from conventional instances in that they have class membership values. Experiments on the Sports and the Iris Flower classification problems are presented to compare the accuracy of the proposed algorithm with those of some other learning algorithms. Experimental results show that the rules derived from our approach are simpler and yield higher accuracy than those from some other learning algorithms. (C) 1999 Elsevier Science B.V. All rights reserved.
A generalized version space learning algorithm for noisy and uncertain data
(1997-03-01)
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
This paper generalizes the learning strategy of version space to manage noisy and uncertain training data. A new learning algorithm is proposed that consists of two main phases: searching and pruning. The searching phase generates and collects possible candidates into a large set; the pruning phase then prunes this set according to various criteria to find a maximally consistent version space. When the training instances cannot completely be classified, the proposed learning algorithm can make a trade-off between including positive training instances and excluding negative ones according to the requirements of different application domains. Furthermore, suitable pruning parameters are chosen according to a given time limit, so the algorithm can also make a trade-off between time complexity and accuracy. The proposed learning algorithm is then a flexible and efficient induction method that makes the version space learning strategy more practical.
A genetics-based approach to knowledge integration and refinement
(2001-01-01)
Wang, CH
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
In this paper, we propose a genetics-based knowledge integration approach to integrate multiple rule sets into a central rule set. The proposed approach consists of two phases: knowledge encoding and knowledge integrating. In the encoding phase, each knowledge input is translated and expressed as a rule set, and then encoded as a bit string. The combined bit strings form an initial knowledge population, which is then ready for integrating. In the knowledge integration phase, a genetic algorithm generates an optimal or nearly optimal rule set from these initial knowledge inputs. Furthermore, a rule-refinement scheme is proposed to refine inference rules via interaction with the environment. Experiments on diagnosing brain tumors were carried out to compare the accuracy of a rule set generated by the proposed approach with that of initial rule sets derived from different groups of experts or induced by means of various machine learning techniques. Results show that the rule set derived using the proposed approach is much more accurate than each initial rule set on its own.
A new hybrid learning algorithm for non-linear boundaries
(1998-06-01)
Wang, CH
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
In this paper, we propose a new hybrid learning algorithm, ETNC, which incorporates the popular decision-tree learning algorithm ASSISTANT into a modified three-layer back-propagation learning method to construct an entropy-tree net classifier. The new Iearning algorithm also adopts a tree-pruning mechanism to avoid overfitting problems. The new algorithm decreases both the tree size and error rate, especially for complex classification problems. Furthermore, it is not necessary for users to lay out the structure of a tree net in advance; instead, the structure is automatically constructed in the tree-growing process. Finally, the results of experiments in diagnosing brain tumors and classifying sugar canes are described to compare the proposed algorithm with two other learning methods, the back-propagation learning algorithm and ASSISTANT, in terms of accuracy, knowledge complexity and learning speed. Experimental results show that the proposed learning algorithm can provide a good trade-off between accuracy, knowledge complexity and learning speed.
A new probabilistic induction method
(1997-02-01)
Hou, RH
;
Hong, TP
;
Tseng, SS
;
Kuo, SY
;
資訊工程學系
Knowledge acquisition by interviewing a domain expert is one of the most problematic aspects of the development of expert systems. As an alternative, methods for inducing concept descriptions from examples have proven useful in eliminating this bottleneck. In this paper, we propose a probabilistic induction method (PIM), which is an improvement of the Chan and Wong method, for detecting relevant patterns implicit in a given data set. PIM uses the technique of residual analysis and several heuristics to effectively detect complex relevant patterns and to avoid the problem of combinatorial explosion. A reasonable trade-off between the induction time and the classification ratio is achieved. Moreover, PIM quickly classifies unknown objects using classification rules converted from the positively relevant patterns detected. Three experiments are conducted to confirm the validity of PIM.
A new probabilistic induction method
(1997-02-01)
Hou, RH
;
Hong, TP
;
Tseng, SS
;
Kuo, SY
;
資訊工程學系
Knowledge acquisition by interviewing a domain expert is one of the most problematic aspects of the development of expert systems. As an alternative, methods for inducing concept descriptions from examples have proven useful in eliminating this bottleneck. In this paper, we propose a probabilistic induction method (PIM), which is an improvement of the Chan and Wong method, for detecting relevant patterns implicit in a given data set. PIM uses the technique of residual analysis and several heuristics to effectively detect complex relevant patterns and to avoid the problem of combinatorial explosion. A reasonable trade-off between the induction time and the classification ratio is achieved. Moreover, PIM quickly classifies unknown objects using classification rules converted from the positively relevant patterns detected. Three experiments are conducted to confirm the validity of PIM.
A parallelized indexing method for large-scale case-based reasoning
(2002-08-01)
Chen, WC
;
Tseng, SS
;
Chang, LP
;
Hong, TP
;
Jiang, MF
Case-based reasoning (CBR) is a problem-solving methodology commonly seen in artificial intelligence. It can correctly take advantage of the situations and methods in former cases to find out suitable solutions for new problems. CBR must accurately retrieve similar prior cases for getting a good performance. In the past, many researchers proposed useful technologies to handle this problem. However, the performance of retrieving similar cases may be greatly influenced by the number of cases. In this paper, the performance issue of large-scale CBR is discussed and a parallelized indexing architecture is then proposed for efficiently retrieving similar cases in large-scale CBR. Several algorithms for implementing the proposed architecture are also described. Some experiments are made and the results show the efficiency of the proposed method. (C) 2002 Elsevier Science Ltd. All rights reserved.
A three-phased online association rule mining approach for diverse mining requests
(2004)
Wang, CY
;
Tseng, SS
;
Hong, TP
;
Chu, YS
;
資訊工程學系
In the past, most incremental mining and online mining algorithms considered finding the set of association rules or patterns consistent with the entire set of data inserted so far. Users can not easily obtain the results from their only interested portion of data. For providing ad-hoc, query-driven and online mining supports, we first propose a relation called multidimensional pattern relation to structurally and systematically store the context information and the mining information for later analysis. Each tuple in the relation comes from an inserted dataset in the database. This concept is similar to the construction of a data warehouse for OLAP. However, unlike the summarized information of fact attributes in a data warehouse, the mined patterns in the multidimensional pattern relation can not be directly aggregated to satisfy, users' mining requests. We then develop an online mining approach called Three-phased Online Association Rule Mining (TOARM) based on the proposed multidimensional pattern relation to support online generation of association rules under multidimensional considerations. Experiments for both homogeneous and heterogeneous datasets are made, with results showing the effectiveness of the proposed approach.
Automatically integrating multiple rule sets in a distributed-knowledge environment
(1998-08-01)
Wang, CH
;
Hong, TP
;
Tseng, SS
;
Liao, CM
;
資訊工程學系
In this paper, an actual knowledge application is made by means of evolution paradigms in terms of knowledge acquisition. rin automatic knowledge integration approach in a distributed-knowledge environment is thus proposed to integrate multiple rule sets into a single effective rule set. The proposed approach consists of two phases: knowledge encoding and knowledge integration. In the encoding phase, each knowledge input is translated and expressed as a rule set, then encoded as a bit string. The combined bit strings from multiple knowledge inputs form an initial knowledge population, which is then ready for integration. In the knowledge integration phase, a genetic search technique generates an optimal or nearly optimal rule set from these initial knowledge-input strings. Finally, experimental results from diagnosis of brain tumors show that the rule set derived by the proposed approach is much more accurate than each initial rule set.
FILSMR: A fuzzy inductive learning strategy for modular rules
(1997)
Wang, CH
;
Liu, JF
;
Hong, TP
;
Tseng, SS
;
交大名義發表
In real applications, data provided to a learning system usually contain linguistic information which greatly influences concept descriptions derived by conventional inductive learning methods. The design of learning methods to learn concept descriptions in linguistic environments is thus very important. In this paper, we apply fuzzy set concepts to machine learning to solve this problem. A fuzzy learning algorithm based on the maximum information gain is proposed to manage linguistic information. Experiments on the sport classification problem are to demonstrate the effectiveness of the proposed algorithm. Experimental results show that the rules derived from our approach are simpler and yields high accuracy.
Flexible online association rule mining based on multidimensional pattern relations
(2006-06-22)
Wang, CY
;
Tseng, SS
;
Hong, TP
;
資訊工程學系
;
Department of Computer Science
Most incremental mining and online mining algorithms concentrate on finding association rules or patterns consistent with entire current sets of data. Users cannot easily obtain results from only interesting portion of data. This may prevent the usage of mining from online decision support for multidimensional data. To provide ad-hoc, query-driven, and online mining support, we first propose a relation called the multidimensional pattern relation to structurally and systematically store context and mining information for later analysis. Each tuple in the relation comes from an inserted dataset in the database. We then develop an online mining approach called three-phase online association rule mining (TOARM) based on this proposed multidimensional pattern relation to support online generation of association rules under multidimensional considerations. The TOARM approach consists of three phases during which final sets of patterns satisfying various mining requests are found. It first selects and integrates related mining information in the multidimensional pattern relation, and then if necessary, re-processes itemsets without sufficient information against the underlying datasets. Some implementation considerations for the algorithm are also stated in detail. Experiments on homogeneous and heterogeneous datasets were made and the results show the effectiveness of the proposed approach. (c) 2005 Elsevier Inc. All rights reserved.
Fuzzy inductive learning strategies
(2003-03-01)
Wang, CH
;
Tsai, CR
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
In real applications, data provided to a learning system usually contain linguistic information which greatly influences concept descriptions derived by conventional inductive learning methods. Design of learning methods for working with vague data is thus very important. In this paper, we apply fuzzy set concepts to machine learning to solve this problem. A fuzzy learning algorithm based on the AQR learning strategy is proposed to manage linguistic information. The proposed learning algorithm generates fuzzy linguistic rules from "soft" instances. Experiments on the Sports and the Iris Flower classification problems are presented to compare the accuracy of the proposed algorithm with those of some other learning algorithms. Experimental results show that the rules derived from our approach are simpler and yield higher accuracy than those from some other learning algorithms.
Improved negative-border online mining approaches
(2006)
Wang, CY
;
Tseng, SS
;
Hong, TP
;
資訊工程學系
;
Department of Computer Science
In the past, we proposed an extended multidimensional pattern relation (EMPR) to structurally and systematically store previously mining information for each inserted block of data, and designed a negative-border online mining (NOM) approach to provide ad-hoc, query-driven and online mining supports. In this paper, we try to use appropriate data structures and design efficient algorithms to improve the performance of the NOM approach. The lattice data structure is utilized to organize and maintain all candidate itemsets such that the candidate itemsets with the same proper subsets can be considered at the same time. The derived lattice-based NOM (LNOM) approach will require only one scan of the itemsets stored in EMPR, thus saving much computation time. In addition, a hashing technique is used to further improve the performance of the NOM approach since many itemsets stored in EMPR may be useless for calculating the counts of candidates. At last, experimental results show the effect of the improved NOM approaches.
Inductive learning from fuzzy examples
(1996)
Wang, CH
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
Integrating membership functions and fuzzy rule sets from multiple knowledge sources
(2000-05-16)
Wang, CH
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
Ln this paper, we propose a GA-based fuzzy knowledge-integration framework that can simultaneously integrate multiple fuzzy rule sets and their membership function sets. The proposed two-phase approach includes fuzzy knowledge encoding and fuzzy knowledge integration. In the encoding phase, each fuzzy rule set with its associated membership functions is first transformed into an intermediary representation, and further encoded as a string. The combined strings form an initial knowledge population which is then ready for integration. In the knowledge-integration phase. a genetic algorithm is used to generate an optimal or nearly optimal set of fuzzy rules and membership functions from the initial knowledge population. The hepatitis diagnostic problem was used to show the performance of the proposed knowledge-integration approach. Results show that the fuzzy knowledge-base resulting from using our approach performs better than every individual knowledge base. (C) 2000 Elsevier Science B.V. All rights reserved.
Learning concepts by arranging appropriate training order
(2001-08-01)
Hsu, YT
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
Machine learning has been proven useful for solving the bottlenecks in building expert systems. Noise in the training instances will, however, confuse a learning mechanism. Two main steps are adopted here to solve this problem. The first step is to appropriately arrange the training order of the instances. It is well known from Psychology that different orders of presentation of the same set of training instances to a human may cause different learning results. This idea is used here for machine learning and an order arrangement scheme is proposed. The second step is to modify a conventional noise-free learning algorithm, thus making it suitable for noisy environment. The generalized version space learning algorithm is then adopted to process the training instances for deriving good concepts. Finally, experiments on the Iris Flower problem show that the new scheme can produce a good training order, allowing the generalized version space algorithm to have a satisfactory learning result.
Machine learning by imitating human learning
(1996-05-01)
Chang, KC
;
Hong, TP
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
Learning general concepts in imperfect environments is difficult since training instances often include noisy data, inconclusive data, incomplete data, unknown attributes, unknown attribute values and other barriers to effective learning. It is well known that people can learn effectively in imperfect environments, and can manage to process very large amounts of data. Imitating human learning behavior therefore provides a useful model for machine learning in real-world applications. This paper proposes a new, more effective way to represent imperfect training instances and rules, and based on the new representation, a Human-Like Learning (HULL) algorithm for incrementally learning concepts well in imperfect training environments. Several examples are given to make the algorithm clearer. Finally, experimental results are presented that show the proposed learning algorithm works well in imperfect learning environments.
Maintenance of sequential patterns for record deletion
(2001)
Wang, CY
;
Hong, TP
;
Tseng, SS
;
交大名義發表
;
National Chiao Tung University
In the past, we proposed an incremental mining algorithm for maintenance of sequential patterns based on the concept of pre-large sequences as new records were inserted. In this paper, we attempt to apply the concept of pre-large sequences to maintain sequential patterns as records are deleted. Pre-large sequences are defined by a lower support threshold and an upper support threshold. They, act as buffers to avoid the movements of sequential patterns directly from large to small and vice-versa. Our proposed algorithm does not require rescanning original databases until the accumulative amount of deleted customer sequences exceeds a safety bound, which depends on database size. As databases grow larger, the numbers of deleted customer sequences allowed before database rescanning is required also grow. The proposed approach is thus efficient for a large database.
Maintenance of sequential patterns for record modification using pre-large sequences
(2002)
Wang, CY
;
Hong, TP
;
Tseng, SS
;
交大名義發表
;
National Chiao Tung University
In the past, we proposed incremental mining algorithms for maintenance of sequential patterns based on the concept of pre-large sequences as records were inserted or deleted Although maintenance of sequential patterns for record modification can be performed by usage of the deletion procedure and then the insertion procedure, twice computation time of a single procedure is needed In this paper, we thus attempt to apply the concept of pre-large sequences to maintain sequential patterns as records are modified. The proposed algorithm does not require rescanning original databases until the accumulative amount of modified customer sequences exceeds a safety bound derived by pre-large concept. As databases grow larger, the numbers of modified customer sequences allowed before database rescanning is required also grow.