Improving the performance of neural networks in classification using fuzzy linear regression

Cheng, Chun Hung; Low, Boon Toh; Chan, Pak-Kei; Motwani, Jaideep

doi:10.1016/s0957-4174(00)00059-2

Cited by 5 publications

(3 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A corollary result from the findings of Table 1 is that perhaps two distinct models are required for accurately categorizing this two-group classification problem. Cheng et al (2001) and others (Walczak 1998) have shown that standard categorization problems may necessitate multiple models (two models for the case of a two-group classification problem) to improve classification accuracy. Thus, if in fact two unique models effectively predict the continuation versus discontinuation behavior of ISP service subscribers, then a means for combining the two appropriate fuzzy ARTMAP neural network models must be developed.…”

Section: A Multiple Agent (Modular) Neural Network Approach To Predicmentioning

confidence: 99%

Modeling online service discontinuation with nonparametric agents

Walczak

Parthasarathy

2005

ISeB

View full text Add to dashboard Cite

The internet and world wide web are an increasingly important resource, both as a market and as an information source, to both individual users and business entities. An estimated 120 million active web users exist in the United States alone. Access to these electronic marketplaces and information sources is accomplished through either a direct internet connection or through a service provider. Internet service providers (ISPs) enable internet and web access for most of these users either via dial-up modems (62.2 percent), or DSL connections (17 percent). Customers of ISPs frequently switch or discontinue service. The model selection perspective is used to extend previous work in this area through the development of a multi-agent system with neural network wrappers. The nonparametric (neural network) agents identify over 92 percent of those users that either stop or change service, which is a 15 percent increase over previous models.

show abstract

Section: A Multiple Agent (Modular) Neural Network Approach To Predicmentioning

confidence: 99%

Modeling online service discontinuation with nonparametric agents

Walczak

Parthasarathy

2005

ISeB

View full text Add to dashboard Cite

show abstract

“…Several methods have been used, including the similar data selective learning method (Peng et al, 1992) and the correlation-coefficient-based similar data selection method (Shimodaira, 1996). Moreover, Cheng et al (2001) demonstrated a method which separated data in a set into good and bad data sets through fuzzy linear regression. Deco and colleagues (1997) present a non-parametric data selection approach for detecting nonlinear statistical dependences in a non-stationary time series.…”

Section: Neural Network For Financial Data Seriesmentioning

confidence: 99%

“…This means that a more homogeneous training set promises a higher performance of neural networks. Cheng et al (2001) propose a way to improve the accuracy of the neural network by separating training data using the fuzzy regression method for classification.…”

Section: Selecting Learning Samplesmentioning

confidence: 99%

Integrated approach of cognitive maps and neural networks using qualitative information on the World Wide Web: the KBNMiner

Hong

Han

2004

Expert Systems

View full text Add to dashboard Cite

In this paper, we propose a method for integrating cognitive maps and neural networks to gain competitive advantage using qualitative information acquired from news information on the World Wide Web. We have developed the KBNMiner, which is designed to represent the knowledge of domain experts with cognitive maps, to search and retrieve news information on the Internet according to the knowledge and to apply the information to a neural network model. In addition, we investigate ways to train neural networks more effectively by separating the learning data into two groups on the basis of event information acquired from news information. To validate our proposed method, we applied 180,000 news articles to the KBNMiner. The experimental results are found to support our proposed method through tenfold cross-validation.

show abstract