Knowldege Discovery

The primary objective of the current research was to explore how the Knowledge Management (KM) processes of Deposit Money Banks (DMBs) in Makurdi Metropolis contribute to their Sustainable Competitive Advantage (SCA). Additionally, this study determined to evaluate how innovation helps the banks to retain their competitive advantage. A total of 265 valid samples were collected from a population of 14 DMBs and 933 respondents were selected from these DMBs. The data was analysed using Structural Equation Modelling (SEM). While testing the assumptions, hypotheses 1-3 supported the proposition that KM dimensions; acquisition (β = .32, p <.05), utilization (β =.26, p <.05), and storage (β = .24, p <.05) were directly related to SCA. No evidence was found for the full mediating role of innovation in relation to KM dimensions and SCA, as demonstrated by the data. However, innovation does play a role in certain aspects of KM by demonstrating that acquisition (β=.69, p <.05), utilisation (β = .21, p <.05), and storage (β = .18, p <.05) are positively associated with innovation in DMBs. The findings indicated that managers who responded to KM were more innovative. The importance of DMBs lies in the recognition that knowledge is not a simple concept. A commitment to new discoveries and to new understandings is not sufficient. Enterprises need to pursue knowledge that enables them to add value.

I would like to express my deep gratitude to my supervisor, Prof. Ho Tu Bao, for providing me with kindly helps, supervision and motivation throughout the course of this work. His insight and breadth of knowledge have been invaluable to my training as a researcher. Without his care, supervision and friendship I would not be able to complete this work. I am also grateful to Dr. Nguyen Trong Dung for his comments and suggestions during the work. My thank also goes to members of the Knowledge Creation Laboratory, especially to Mr. Saito Akinori, Ms. Kawasaki Saori, for providing their helps, a friendly and enjoyable environment while studying here; Mr. Nghiem Anh Tuan for his valuable discussions and comments. I would like to express my appreciation to the Ministry of Education, Culture, Sports, Science, and Technology of Japan for providing me the scholarship for pursuing the master course at JAIST, and for the financial support for attending several conferences. Finally, I am indebted to my parents for their forever affection, patience, and constant encouragement.

Many of the existing classifiers cannot deal with exceptions, which are not to be ignored in real life. In this paper, an exception-tolerant methodology is proposed based on each of the following three popular algorithms for multi-class classification problems: C4.5, PRISM and RISE. The performance of each algorithm was improved by converting the outputs into the format of RISE induced rules, applying bagging and boosting techniques, adding exceptions with a default rule, and excluding inefficient rules. The improved versions of C4.5, PRISM and RISE are named as OE 2-C4.5, OE 2-PRISM and OE 2-RISE, respectively. Note that OE 2 stands for Ordering of Efficient Rules and Inclusion of Exceptions. Empirical results show that OE 2-C4.5 and OE 2-RISE significantly outperformed classical C4.5, PRISM and RISE for each dataset. Our methodology provides a reference for improving other weak learners individually or in an ensemble.

Forecasting the full distribution of the number of earthquakes is revealed to be inherently superior to forecasting their mean. Forecasting the full distribution of earthquake numbers is also shown to yield robust projections in the presence of "surprise" large earthquakes, which in the past have strongly deteriorated the scores of existing models. We show this with pseudoprospective experiments on synthetic as well as real data from the Advanced National Seismic System (ANSS) database for California, with earthquakes with magnitude larger than 2.95 that occurred between the period 1971-2016. Our results call in question the testing methodology of the Collaboratory for the study of earthquake predictability (CSEP), which amounts to assuming a Poisson distribution of earthquake numbers, which is known to be a poor representation of the heavy-tailed distribution of earthquake numbers. Using a spatially varying ETAS model, we demonstrate a remarkable stability of the forecasting performance, when using the full distribution of earthquake numbers for the forecasts, even in the presence of large earthquakes such as Mw 7.1 Hector Mine, Mw 7.2 El Mayor-Cucapah, Mw 6.6 Sam Simeon earthquakes, or in the presence of intense swarm activity in Northwest Nevada in 2014.

This repon shortly describes calculation method related to guiding propenies of the microstructure fibres. For thc modelling of wave propagation in such fibres in linear and nonlinear regimes the finite-difference vectorial-beampropagation method (FD-VBMP) has been chosen. This method offers high accuracy and allows investigating longitudinally varying structules oI propagation ofoptical waves with amplitudes varying under the effect ofnonlinęaritr and loss. ln order to take into account the effects related to the finite fibre dimensions the transparent boundarl conditions (TBC) was used. The Split-Step Local Error method is implemented for correct estimation ofthe influencc.'i nonlinearity on the wave propagation in microstructure fibres. On the basis ofthis method the complete algorithm for thc numerical simulation of the wave propagation in microstructure fibres under the effect of nonlinearity has becn developed. The method allows optimising the structure of photonic crystal fibres (PCF) for maximising or minimisin-r nonlinear ęffects. lnvestigation of the propenies of a wavelength conveltel based on a miclostructue fibre was carri.'.l out using simulation results in the nonlinear case. Modification of the method for optimisation of connection betwecr microstructure fibres and conventional fibres is also considered.

Knowledge Acquisition is a major problem in developing ES, it is costly and time consuming. An alternate to the acquisition of knowledge through the dialog with an expert is to convert a history of previous cases into a set of IF-THEN rules. DSS are considered as rich sources of such cases and can be integrated with ES to build an integrated decision expert system with automatic knowledge acquisition capability. In this paper, three contributions have been presented: (1) A framework for integrating DSS and ES has been proposed, (2) In this framework, a new inductive learning algorithm (called ERIC) for extracting rules from DSS has been presented Experiments showed that the results of this algorithm are better than the results of some well-known algorithms in the field of inductive learning, and (3) A DSS and an ES have been built and integrated with each other using the proposed framework.

Most concept learning algorithms are conjunctive algorithms, i.e. generate production rules that include AND-operators only. This paper examines the induction of disjunctive concepts or descriptions. We present an algorithm, called DCL, for disjunctive concept learning that partitions the training data according to class descriptions. DCL generates production rules with AND/OR-operators from a set of training examples. This approach is particularly useful for creating multiple decision boundaries. We also describe application of DCL to a range of training sets with different number of attributes and classes. The results obtained show that DCL can produce fewer number of rules than most other algorithms used for inductive concept learning, and also can classify considerably more unseen examples than conjunctive algorithms.

In the feature subset selection problem, a learning algorithm is faced with the problem of selecting a relevant subset of features upon which to focus its attention, while ignoring the rest. To achieve the best possible performance with a particular learning algorithm on a particular domain, a feature subset selection method should consider how the algorithm and the training data interact. We explore the relation between optimal feature subset selection and relevance. Our wrapper method searches for an optimal feature subset tailored to a particular algorithm and a domain. We study the strengths and weaknesses of the wrapper approach and show improvements over the original design. We compare the wrapper approach to induction without feature subset selection and to Relief, a lter-based approach to feature subset selection. Signi cant improvement in accuracy on real problems is achieved for the two families of induction algorithms used: decision trees and Naive-Bayes.

In this paper we describe the ILA-2 rule induction algorithm which is the improved version of a novel inductive learning algorithm, ILA. We first outline the basic algorithm ILA, and then present how the algorithm is improved using a new evaluation metric that handles uncertainty in the data. By using a new soft computing metric, users can reflect their preferences through a penalty factor to control the performance of the algorithm. ILA has also a faster pass criteria feature which reduces the processing time without sacrificing much from the accuracy that is not available in basic ILA.
We experimentally show that the performance of ILA-2 is comparable to that of well-known inductive learning algorithms, namely CN2, OC1, ID3 and C4.5.

In this paper we describe the ILA-2 rule induction algorithm which is the improved version of a novel inductive learning algorithm, ILA. We first outline the basic algorithm ILA, and then present how the algorithm is improved using a new evaluation metric that handles uncertainty in the data. By using a new soft computing metric, users can reflect their preferences through a penalty factor to control the performance of the algorithm. ILA has also a faster pass criteria feature which reduces the processing time without sacrificing much from the accuracy that is not available in basic ILA.

In this paper we describe the ILA-2 rule induction algorithm which is the improved version of a novel inductive learning algorithm, ILA. We first outline the basic algorithm ILA, and then present how the algorithm is improved using a new evaluation metric that handles uncertainty in the data. By using a new soft computing metric, users can reflect their preferences through a penalty factor to control the performance of the algorithm. ILA has also a faster pass criteria feature which reduces the processing time without sacrificing much from the accuracy that is not available in basic ILA.

In this paper we describe the ILA-2 rule induction algorithm which is the improved version of a novel inductive learning algorithm, ILA. We first outline the basic algorithm ILA, and then present how the algorithm is improved using a new evaluation metric that handles uncertainty in the data. By using a new soft computing metric, users can reflect their preferences through a penalty factor to control the performance of the algorithm. ILA has also a faster pass criteria feature which reduces the processing time without sacrificing much from the accuracy that is not available in basic ILA.

Among decision tree classifiers, Bayesian classifiers,
k-nearest-neighbor classifiers, case-based reasoning, genetic
algorithms, rough sets, and fuzzy logic which are some common
classification methods in data mining, decision tree classifier is
the most commonly used because it is easier to understand and
implement compared to other classification methods. In this
study, performance evaluation of two decision tree algorithm
named ID3 and C4.5is done in order to discover which one is
superior on the basis of accuracy with the help of a software tool
using publicly available datasets.

An effectiv e modeling met ho d o f do main level co nst raint s in t he co nstr aint netw or k for co ncur rent engineer ing ( CE) w as developed. T he domain lev el constr aints w ere analy zed and the fr amew or k of modeling of domain level co nst raint s based o n simulat ion and appro x imate t echnolo gy was g iv en. A n intellig ent r esponse surface methodolo gy ( IRSM ) w as pr opo sed, in w hich artificial intellig ence technolog ies ar e int ro duced into the optimizatio n pr ocess. T he desig n o f cr ank and connect ing r od in the V 6 eng ine as ex ample was given to sho w the validity of the modeling met ho d. Key words: co nstr aint netw or k; do main level co nst raints; simulation; int ellig ent response sur face met ho do log y ( I RSM ) CLC number: T P 391. 4

In this paper we describe the ILA-2 rule induction algorithm which is the improved version of a novel inductive learning algorithm, ILA. We first outline the basic algorithm ILA, and then present how the algorithm is improved using a new evaluation metric that handles uncertainty in the data. By using a new soft computing metric, users can reflect their preferences through a penalty factor to control the performance of the algorithm. ILA has also a faster pass criteria feature which reduces the processing time without sacrificing much from the accuracy that is not available in basic ILA.

ABSTRACT Feature selection is defined as a problem to find a minimum set of M features for an inductive algorithm to achieve the highest predictive accuracy from the data described by the original Ar features where.!/<. Y. A probabilistic wrapper model is proposed as another method besides the exhaustive search and the heuristic approach. The aim of this model is to avoid local minima and exhaustive search. The highest predictive accuracy is the criterion in search of the smallest M.