Time-series analysis with a hybrid Box-Jenkins ARIMA

Time-series analysis is important to a wide range of disciplines transcending both the physical and social sciences for proactive policy decisions. Statistical models have sound theoretical basis and have been successfully used in a number of problem domains in time series forecasting. Due to power and flexibility, Box-Jenkins ARIMA model has gained enormous popularity in many areas and research practice for the last three decades. More recently, the neural networks have been shown to be a promising alternative tool for modeling and forecasting owing to their ability to capture the nonlinearity in the data. However, despite the popularity and the superiority of ARIMA and ANN models, the empirical forecasting performance has been rather mixed so that no single method is best in every situation. In this study, a hybrid ARIMA and neural networks model to time series forecasting is proposed. The basic idea behind the model combination is to use each model’s unique features to capture different patterns in the data. With three real data sets, empirical results evidently show that the hybrid model outperforms ARIMA and ANN model noticeably in terms of forecasting accuracy used in isolation.

Airflow adjustment and minimization of the air power of ventilation network

It is often required to know which roadway （adjustment roadway） resistances and how much values of the resis- tances should be changed to make the airflow rates in roadways （target roadways） to certain required values in the practice of mine ventilation. In this case, the airflow rates of the target roadways and the resistances of the roadways other than the ad- justment roadways are the given conditions and the resistances of the adjustment roadways are the solutions to be found. No straightforward method to solve the problem has been found up to now. Therefore, trial and error method using the ventilation network analysis program is utilized to solve the problem so far. The method takes long calculation time and the best answer is not necessarily obtained. The authors newly defined ＂airflow element＂ as an element of the ventilation network analysis. The resistances that satisfy the airflow requirements can be calculated straight forwardly by putting the function of the airflow element into the ventilation network analysis. The air power required for the ventilation can be minimized while meeting the airflow requirements by the advanced application of the method. The authors made the computer program fulfill the method. The program was applied to actual ventilation network and it was found that the method is very practical and the time required for the analysis is short.

SIMULATION AND PREDICTION OF DEBRIS FLOW USING ARTIFICIAL NEURAL NETWORK

Debris flow is one of the most destructive phenomena of natural hazards. Recently, major natural hazard, claiming human lives and assets, is due to debris flow in the world. Several practical methods for forecasting debris flow have been proposed, however, the accuracy of these methods is not high enough for practical use because of the stochastic and non-linear characteristics of debris flow. Artificial neural network has proven to be feasible and useful in developing models for nonlinear systems. On the other hand, predicting the future behavior based on a time series of collected historical data is also an important tool in many scientific applications. In this study we present a three-layer feed-forward neural network model to forecast surge of debris flow according to the time series data collected in the Jiangjia Ravine, situated in north part of Yunnan Province of China. The simulation and prediction of debris flow using the proposed approach shows this model is feasible, however, further studies are needed.

Pattern Specialization Based on Partial Evaluation

Design pattern enables software architecture generality and reusability, but which depresses the high performance. The pattern specialization was built on partial evaluation technology to reduce the overheads of design pattern. The design patterns were classified to extract the common features, and the corresponding pattern specializations were constructed. In the pattern specialization, the optimization opportunities were identified, and the specialization methods and conditions were described. The syntax of binding time analysis was defined, and the semantic depicted the invariant of usage context. The virtual invocation and dispatch were eliminated, which enhances the running efficiency. This pattern specialization is a high-level specialization for improving the performance of software aimed at design level that is orthogonal with the low-level code optimization.

Neural Networks and the Study of Time Series： An Application in Engineering Education

Time series are an important object of study in sciences, engineering and business, especially in cases where it is expected to know, predict and optimize behaviors. In this context, we intend to show the feasibility of using artificial neural networks in the study of several time series in an engineering course, especially those that have no overt behavior or are not able to be modeled mathematically in a simple way and have direct application in the education of future engineers.

Exponential H_∞ filtering analysis for discrete-time switched neural networks with random delays using sojourn probabilities

This paper is concerned with the exponential H_∞ filtering problem for a class of discrete-time switched neural networks with random time-varying delays based on the sojourn-probability-dependent method. Using the average dwell time approach together with the piecewise Lyapunov function technique, sufficient conditions are proposed to guarantee the exponential stability for the switched neural networks with random time-varying delays which are characterized by introducing a Bernoulli stochastic variable.Based on the derived H_∞ performance analysis results, the H_∞ filter design is formulated in terms of Linear Matrix Inequalities(LMIs). Finally, two numerical examples are presented to demonstrate the effectiveness of the proposed design procedure.

Multifractal Detrended Fluctuation Analysis of Interevent Time Series in a Modified OFC Model

We use multifractal detrended fluctuation analysis （MF-DFA） method to investigate the multifractal behavior of the interevent time series in a modified Olami-Feder-Christensen （OFC） earthquake model on assortative scale-free networks. We determine generalized Hurst exponent and singularity spectrum and find that these fluctuations have multifraetal nature. Comparing the MF-DFA results for the original interevent time series with those for shuffled and surrogate series, we conclude that the origin of multifractality is due to both the broadness of probability density function and long-range correlation.