International Journal of Computational Intelligence and Applications

Forecasts of the San Diego and San Francisco S&P/Case-Shiller Home Price Indices through December 2012 are obtained using a multi-agent system that utilizes January, 2002–June, 2011 data. Agents employ genetic programming (GP) and neural networks (NN) in a three-stage process to produce fits and forecasts. First, GP and NN compete to provide independent predictions. In the second stage, they cooperate by fitting the first-stage competitor's residuals. Outputs from the first two stages then become inputs to produce two final GP and NN outputs. The NN output from the third stage using the combined method produces improved forecasts over the 3-stage GP method as well as those produced by either method alone. The proposed methodology serves as an example of how combining more than one estimation/forecasting technique may lead to more accurate forecasts.

This paper introduces a novel numerical stochastic optimization algorithm inspired from the behaviors of cloud in the natural world, which is designated as atmosphere clouds model optimization (ACMO) algorithm. It is tried to simulate the generation behavior, move behavior and spread behavior of cloud in a simple way. The ACMO algorithm has been tested on a set of benchmark functions in comparison with two other evolutionary-based algorithms: particle swarm optimization (PSO) algorithm and genetic algorithm (GA). The results demonstrate that the proposed algorithm has certain advantages in solving multimodal functions, while the PSO algorithm has a better result in terms of convergence accuracy. In conclusion, the ACMO algorithm is an effective method in solving optimization problems.

Machine learning methods, such as neural network (NN) and support vector machine, assume that the training data and the test data are drawn from the same distribution. This assumption may not be satisfied in many real world applications, like long-term financial failure prediction, because the training and test data may each come from different time periods or domains. This paper proposes a novel algorithm known as fuzzy bridged refinement-based domain adaptation to solve the problem of long-term prediction. The algorithm utilizes the fuzzy system and similarity concepts to modify the target instances' labels which were initially predicted by a shift-unaware prediction model. The experiments are performed using three shift-unaware prediction models based on nine different settings including two main situations: (1) there is no labeled instance in the target domain; (2) there are a few labeled instances in the target domain. In these experiments bank failure financial data is used to validate the algorithm. The results demonstrate a significant improvement in the predictive accuracy, particularly in the second situation identified above.

The quality of test data has an important impact on the effect of software testing, so test data generation has always been a key task for finding the potential faults in program code. In structural testing, the primary goal is to cover some kinds of structure elements with some specific inputs. Search-based test data generation provides a rational way to handle this difficult problem. In the past, some well-known meta-heuristic search algorithms have been successfully utilized to solve this issue. In this paper, we introduce a variant of genetic algorithm (GA), called quantum-inspired genetic algorithm (QIGA), to generate the test data with stronger coverage ability. In this new algorithm, the traditional binary bit is replaced by a quantum bit (Q-bit) to enlarge the search space so as to avoid falling into local optimal solution. On the other hand, some other strategies such as quantum rotation gate and catastrophe operation are also used to improve algorithm efficiency and quality of test data. In addition, experimental analysis on eight real-world programs is performed to validate the effectiveness of our method. The results show that QIGA-based method can generate test data with higher coverage in much smaller convergence generations than GA-based method. More importantly, our proposed method is more robust for algorithm parameter change.

Foreign exchange rate (forex) forecasting has been the subject of several rigorous investigations due to its importance in evaluating the benefits and risks of the international business environments. Many methods have been researched with the ultimate goal being to increase the reliability and efficiency of the forecasting method. However as the data are inherently dynamic and complex, the development of accurate forecasting method remains a challenging task if not a formidable one. This paper proposes a new weight of the fuzzy time series model for a daily forecast of the exchange rate market. Through this method, the weights are assigned to the fuzzy relationships based on a probability approach. This can be implemented to carry out the frequently recurring fuzzy logical relationship (FLR) in the fuzzy logical group (FLG). The US dollar to the Malaysian Ringgit (MYR) exchange rates are used as an example and the efficiency of the proposed method is compared with the methods proposed by Yu and Cheng et al. The result shows that the proposed method has enhanced the accuracy and efficiency of the daily exchange rate forecasting opportunities.

In this paper, we present a parallel implementation of the particle swarm optimization (PSO) on graphical processing units (GPU) using CUDA. By fully utilizing the processing power of graphic processors, our implementation (CUDA-PSO) provides a speedup of 167× compared to a sequential implementation on CPU. This speedup is significantly superior to what has been reported in recent papers and is achieved by four optimizations we made to better adapt the parallel algorithm to the specific architecture of the NVIDIA GPU. However, because today's personal computers are usually equipped with a multicore CPU, it may be unfair to compare our CUDA implementation to a sequential one. For this reason, we implemented a parallel PSO for multicore CPUs using MPI (MPI-PSO) and compared its performance against our CUDA-PSO. The execution time of our CUDA-PSO remains 15.8× faster than our MPI-PSO which ran on a high-end 12-core workstation. Moreover, we show with statistical significance that the results obtained using our CUDA-PSO are of equal quality as the results obtained by the sequential PSO or the MPI-PSO. Finally, we use our parallel PSO for real-time harmonic minimization of multilevel power inverters with 20 DC sources while considering the first 100 harmonics and show that our CUDA-PSO is 294× faster than the sequential PSO and 32.5× faster than our parallel MPI-PSO.