Time series prediction techniques have been used in many real-world applications such as financial market prediction, electric utility load forecasting , weather and environmental state prediction, and reliability forecasting. The underlying system models and time series data generating processes are generally complex for these applications and the models for these systems are usually not known a priori. Accurate and unbiased estimation of the time series data produced by these systems cannot always be achieved using well known linear techniques, and thus the estimation process requires more advanced time series prediction algorithms. This paper provides a survey of time series prediction applications using a novel machine learning approach: support vector machines (SVM). The underlying motivation for using SVMs is the ability of this methodology to accurately forecast time series data when the underlying system processes are typically nonlinear, non-stationary and not defined a-priori. SVMs have also been proven to outperform other non-linear techniques including neural-network based non-linear prediction techniques such as multi-layer perceptrons.The ultimate goal is to provide the reader with insight into the applications using SVM for time series prediction, to give a brief tutorial on SVMs for time series prediction, to outline some of the advantages and challenges in using SVMs for time series prediction, and to provide a source for the reader to locate books, technical journals, and other online SVM research resources.

Time Series Prediction Using Support Vector Machines: A Survey

Application of least square support vector machine and multivariate adaptive regression spline models in long term prediction of river water pollution

Covariance matching based adaptive unscented Kalman filter for direct filtering in INS/GNSS integration

A heuristic method for parameter selection in LS-SVM: Application to time series prediction

Accurate heat rate forecasting is very important in ensuring the economic, efficient, and safe operation of a steam turbine unit. The support vector machine (SVM) is a novel tool from the artificial intelligence field that has been successfully applied to heat rate forecasting. The least squares SVM (LS-SVM) is an improved algorithm based on the SVM. LS-SVM has minimal computational complexity and fast calculation. However, traditional LS-SVM, which was established by using offline data samples, can no longer accurately describe the actual system working condition, thereby resulting in problems when directly used in heat rate prediction. In this paper, a heat rate forecasting method based on online LS-SVM, which possesses dynamic prediction functions, is proposed. To avoid blindness and inaccuracy in parameter selection, the gravitational search algorithm (GSA) is used to optimize the regularization parameter @c and the kernel parameter @s^2 of the online LS-SVM modeling. The results confirm the efficiency of the proposed method.

Forecasting of turbine heat rate with online least squares support vector machine based on gravitational search algorithm

A constant velocity model, a constant acceleration model and a coordinated turn model are used for manoeuvring target. However, high location precision of the target could not be obtained with any one of these models. In this paper, an interacting multiple model particle filter (IMMPF) algorithm is proposed to estimate the target location with several models. Besides similar mixing and interaction to those in a traditional interaction multiple model (IMM) estimator, a standard particle filter runs in every model and the number of particles in every model is fixed. Compared through a target location example, the proposed algorithm is better than the traditional IMM.

https://iopscience.iop.org/article/10.1088/0957-0233/17/6/003/pdf

An interacting multiple model particle filter for manoeuvring target location

A marine INS/GPS (inertial navigation system/global positioning system) adaptive navigation system is presented in this paper. The GPS with two antennae providing vessel attitude is selected as the auxiliary system to fuse with INS. The Kalman filter is the most frequently used algorithm in the integrated navigation system, which is capable of estimating INS errors online based on the measured errors between INS and GPS. The conventional Kalman filter (CKF) assumes that the statistics of the noise of each sensor are given. As long as the noise distributions do not change, the Kalman filter will give the optimal estimation. However, the GPS receiver will be disturbed easily and thus temporally changing measurement noise will join into the outputs of GPS, which will lead to performance degradation of the Kalman filter. Many researchers introduce a fuzzy logic control method into innovation-based adaptive estimation Kalman filtering (IAE-AKF) algorithm, and accordingly propose various adaptive Kalman filters. However, how to design the fuzzy logic controller is a very complicated problem, which is still without a convincing solution. A novel IAE-AKF is proposed herein, which is based on the maximum likelihood criterion for the proper computation of the filter innovation covariance and hence of the filter gain. The approach is direct and simple without having to establish fuzzy inference rules. After having deduced the proposed IAE-AKF algorithm theoretically in detail, the approach is tested in the developed INS/GPS integrated marine navigation system. Real field test results show that the adaptive Kalman filter outperforms the CKF with higher accuracy and robustness. It is demonstrated that this proposed approach is a valid solution for the unknown changing measurement noise existing in the Kalman filter.

Study on GPS attitude determination system aided INS using adaptive Kalman filter

Temperature drift is the main source of errors affecting the precision and performance of a dynamically tuned gyroscope (DTG). In this paper, the support vector machine (SVM), a novel learning machine based on statistical learning theory (SLT), is described and applied in the temperature drift modelling and compensation to reduce the influence of temperature variation on the output of the DTG and to enhance its precision. To improve the modelling and compensation capability, wavelet transform (WT) is introduced into the SVM model to eliminate any impactive noises. The real temperature drift data set from the long-term measurement system of a certain DTG is employed to validate the effectiveness of the proposed combination strategy. Moreover, the traditional neural network (NN) approach is also investigated as a comparison with the SVM based method. The modelling and compensation results indicate that the proposed WT-SVM model outperforms the NN and single SVM models, and is feasible and effective in temperature drift modelling and compensation of the DTG.

https://iopscience.iop.org/article/10.1088/0957-0233/18/5/030/pdf

Temperature drift modelling and compensation for a dynamically tuned gyroscope by combining WT and SVM method

In this paper, the support vector machine (SVM), a novel learning machine based on statistical learning theory (SLT), is described and applied in the drift modelling of the dynamically tuned gyroscope (DTG). As a data preprocessing method, accumulated generating operation (AGO) is applied to the SVM for further improving the modelling precision and the learning performance of the drift model. The grey modelling method and RBF neural network are also investigated as a comparison to the SVM and AGO–SVM modelling methods. The modelling results of the real drift data from the long-term measurement system of a DTG indicate that the SVM method is available practically in the modelling of DTG drift and the proposed strategy of combining SVM with AGO is effective in improving the modelling precision and the learning performance.

An AGO?SVM drift modelling method for a dynamically tuned gyroscope

The demand for remote sensing satellites is ever increasing due to their potential applications in many areas such as agriculture, forestry, water resources and so on. One of the key concerns in remote sensing satellite applications is that the accuracy of the attitude and orbit control must be maintained at a high level. In general applications, measurement of the attitude using only one type of sensor is satisfactory. However, in applications where a high precision of attitude control is required, several types of sensor need to be integrated together to ensure high measurement accuracy. In these cases, data fusion is required to process the information collected from different sensors. The least mean square method is commonly used for data fusion. This method is simple to use but not accurate enough. The Kalman filter (KF) is an alternative choice, which produces more accurate results, but a conventional KF fails to ensure error convergence due to limited knowledge of the system's dynamic model and measurement noise. In this paper, a modified KF is proposed to overcome the error divergence problem. The possible decrease in the signal processing results which appears in the classical KF can be avoided. The effectiveness of this approach has been demonstrated through a simulation study of the attitude measurement by the attitude and orbit control system of a remote sensing satellite used by the China Astronautic Board.

Zhihua Jin

Papers

An interacting multiple model particle filter for manoeuvring target location

Study on GPS attitude determination system aided INS using adaptive Kalman filter

Temperature drift modelling and compensation for a dynamically tuned gyroscope by combining WT and SVM method

An AGO?SVM drift modelling method for a dynamically tuned gyroscope

Sensor fusion in remote sensing satellites using a modified Kalman filter