Weigang Zhao

Optimal parameters estimation and input subset for grey model based on chaotic particle swarm optimization algorithm

Optimum prediction is a difficult problem, because there are no optimal models for all forecasting problems. In this paper, the authors attempt to find the high precision prediction for grey forecasting model (GM). Considering that chaotic particle swarm optimization algorithm (CPSO) will not get into local optimum and is easy to implement, the paper develops an approach for grey forecasting model, which is particularly suitable for small sample forecasting, based on chaotic particle swarm optimization and optimal input subset which is a new concept. The input subset of traditional time series consists of the whole original data, but the whole original does not always reflect the internal regularity of time series, so the new optimal subset method is proposed to better reflect the internal characters of time series and improve the prediction precision. The numerical simulation result of financial revenue demonstrates that developed algorithm provides very remarkable results compared to traditional grey forecasting model for small dataset forecasting.

ARIMA Model Estimated by Particle Swarm Optimization Algorithm for Consumer Price Index Forecasting

This paper presents an ARIMA model which uses particle swarm optimization algorithm (PSO) for model estimation. Because the traditional estimation method is complex and may obtain very bad results, PSO which can be implemented with ease and has a powerful optimizing performance is employed to optimize the coefficients of ARIMA. In recent years, inflation and deflation plague the world moreover the consumer price index (CPI) which is a measure of the average price of consumer goods and services purchased by households is usually observed as an important indicator of the level of inflation, so the forecast of CPI has been focused on by both scientific community and relevant authorities. Furthermore, taking the forecast of CPI as a case, we illustrate the improvement of accuracy and efficiency of the new method and the result shows it is predominant in forecasting.

Application of GMDH to Short-Term Load Forecasting

Daily power load forecasting plays a significant role in electrical power system operation and planning. Therefore, it is necessary to find automatic interrelations of data and select the optimal structure of model. However, obtaining high accuracy by using single model for short-term load forecasting (STLF) is not easy. In this paper, Group Method of Data Handling (GMDH) is applied to forecast electric load demand of New South Wales (NSW) in Australia from January 17, 2009 to January 18, 2009. Compared with outcomes obtained by ARIMA, we demonstrate that GMDH is a better method for STLF.

Higher-Order Multivariate Markov Chains Based on Particle Swarm Optimization Algorithm for Air Pollution Forecasting

This paper presents a higher-order multivariate Markov chain model combined with particle swarm optimization algorithm. Due to some deficiencies, such as only considering the maximum probability while ignoring the effect of the other probabilities, the traditional method of probability distribution has been replaced by the level characteristics value of fuzzy set theory; further more Particle swarm optimization algorithm has been employed to optimize the coefficient of level characteristics value. In recent years, air pollution acutely aggravates chronic diseases in mankind, such as sulfur dioxide pollution which plays a most important role in acid rain. In order to confront air pollution problems and to plan abatement strategies, both the scientific community and the relevant authorities have focused on monitoring and analyzing the atmospheric pollutants concentration. Taking the forecast of air pollutants as a case, we illustrate the improvement of accuracy and efficiency of the new method and the result shows the new method is predominant in forecasting of multivariate and non-linear data.

A Rough Set Based PSO-BPNN Model for Air Pollution Forecasting

Based on rough set theory, a multilayer back propagation neural network (BPNN) whose parameters will be trained and optimized by particle swarm optimization (PSO) is presented here. Making use of the intelligence of RS in knowledge acquisition aspect, this method carries out a pretreatment on the BPNN data, extracts the regulation from large amount of original data, predigests the nerve basics in neural networks, facilitate the neural networks structure, then employ PSO to the weight parameter and finally improve systematic speed and forecasting accuracy. After data pretreatment and attribute reduction by employing RS theory, the noise data and weak interdependency term are eliminated, so the influences during the initialization, study and training process are avoided, and then the weight parameters of each nerve cell have been optimized through PSO, as a result the accuracy of predictions is developed and proved by the evidence of forecasting with time series from the concentration of air pollutant.