2007, Volume XII, nº 2
Permanent URI for this collection
Browse
Recent Submissions
- Using an integrated fuzzy inference system and artificial neural network to forecast daily dischargePublication . Chang-Shian, Chen; You-Da, Jhong; Chao-Hsien, YehGiven the nonlinearity and uncertainty in the rainfall-runoff process, estimating or predicting hydrologic data often encounters tremendous difficulty. This study applied fuzzy theory to create a daily flow forecasting modeL To improve the time-consuming definition process of membership function, which is usually concluded by a trial-and-error approach, this study designated the membership function by artificial neural network {ANN} with either a supervised or unsupervised learning procedure. The supervised learning was processed by the adaptive network based fuzzy inference system {ANFIS}, while the unsupervised learning was created by fuzzy and self-organizing map {SOMFIS}. The results indicate that the ANFIS method under increment flow data could provide more precise results for daily flow forecasting.
- Dynamic aperiodic neural network for time series predictionPublication . Chiu-Che, TsengThere are many things that humans find easy to do that computers are currently unable to do. Tasks such as visual pattern manipulating objects by touch, and navigating in a complex world are easy for humans. Yet, despite decades of research, we have no viable algorithms for performing these and other cognitive functions on a computer. In this study, we used a bio-inspired neural network called a KA set neural network to perform a time series predictive task. The results from our experiments showed that the predictive accuracy with this method was better in most markets than results obtained using a random walk method.
- Applying recognition of emotions in speech to extend the impact of brand slogan researchPublication . Chien, Charles S.; Wan-Chen, Wang; Moutinho, Luiz; Cheng, Yun-Maw; Pao, Tsang-Long; Yu-Te, Chen; Jun-Heng, YehHow brand slogans can influence and change the consumers' perception of image of products has been a topic of great interest to marketers. However, it is a non-trivial task to evaluate how brand slogans affect their customers' emotions and how the emotions influence the customers' perceptions of brand images. In this paper we demonstrate the Slogan Validator to evaluate brand slogans by analyzing the speech signals from customers' voiced slogans. It is arguably the first speech signal based analysis of brand slogans. Our intention was to evaluate whether the signal-based emotion recognition technique can complement the traditional research methodologies, such as survey research method dealing with self reported measurements, phenomenological research based on physiological measures, and semi-structured interviews, in order to increase the overall effectiveness of advertising copy strategy. The preliminary results of the experiment show that the Slogan Validator yields high consistency with the participants' actual perceptions of the brand slogans chosen for this research.
- Electricity market price forecasting by grid computing optimizing artificial neural networksPublication . Niimura, T.; Ozawa, K.; Sakamoto, N.This paper presents a grid computing approach to parallel-process a neural network time-series model for forecasting electricity market prices. A grid computing environment introduced in a university computing laboratory provides access to otherwise underused computing resources.The grid computing of the neural network model not only processes several times faster than a single iterative process, but also provides chances of improving forecasting accuracy. Results of numerical tests using real market data on twenty grid-connected PCs are reported.
- Agent-based modeling to investigate the disposition effect in financial marketsPublication . Lin, Shi-Woei; Huang, Hui-LungOne of the behavioral patterns that deviate from what is predicted by traditional financial theories is the disposition effect. Although most empirical studies have reported a significant disposition effect, researchers have yet to conduct a conclusive test of thiseffect because a competing hypothesis or confounding effects might explain the documented significance. Thus, we use the tools of computational intelligence, instead of empirical approaches, to explore market behavior. In particular, we allow agents with different investment strategies to interact and to compete with each other in an artificial futures market. We found that the S-shaped value curve proposed by prospect theory may be one of the causes of the observed behavior of the disposition effect. However, rational expectation such as short-term mean reversion can even be more decisive.