: This thesis applies neural network feature selection techniques to multivariate time series data to improve prediction of a target time series. Two approaches to feature selection are used. First, a subset enumeration method is used to determine which financial indicators are most useful for aiding in prediction of the S&P 500 futures daily price. The candidate indicators evaluated include RSI, Stochastics and several moving averages. Results indicate that the Stochastics and RSI indicators result in better prediction results than the moving averages. The second approach to feature selection is calculation of individual saliency metrics. A new decision boundary-based individual saliency metric, and a classifier independent saliency metric are developed and tested. Ruck's saliency metric, the decision boundary based saliency metric, and the classifier independent saliency metric are compared for a data set consisting of the RSI and Stochastics indicators as well as delayed closing price values. The decision based metric and the Ruck metric results are similar, but the classifier independent metric agrees with neither of the other metrics. The nine most salient features, determined by the decision boundary based metric, are used to train a neural network and the results are presented and compared to other published results. (AN)

/pdf/nonlinear-time-series-analysis-1j3a35n4y3.pdf

Nonlinear Time Series Analysis.

random data analysis and measurement procedures is available in our digital library an online access to it is set as public so you can get it instantly. Our book servers spans in multiple countries, allowing you to get the most less latency time to download any of our books like this one. Merely said, the random data analysis and measurement procedures is universally compatible with any devices to read.

Random Data Analysis And Measurement Procedures

In this paper a novel feature extraction technique for micro-Doppler classification and its real time implementation using SVM on an embedded low-cost DSP are presented. The effectiveness of the proposed technique is improved through the exploitation of the outlier rejection capabilities of the Robust PCA in place of the classic PCA.

/pdf/robust-pca-micro-doppler-classification-using-svm-on-4ttrjz0m05.pdf

Robust PCA micro-doppler classification using SVM on embedded systems

Target motions, other than the main bulk translation of the target, induce Doppler modulations around the main Doppler shift that form what is commonly called a target micro-Doppler signature. Radar micro-Doppler signatures are generally both target and action specific and hence can be used to classify and recognise targets as well as to identify possible threats. In recent years, research into the use of micro-Doppler signatures for target classification to address many defence and security challenges has been of increasing interest. In this article, we present a review of the work published in the last 10 years on emerging applications of radar target analysis using micro-Doppler signatures. Specifically we review micro-Doppler target signatures in bistatic SAR and ISAR, through-the-wall radar and ultrasound radar. This article has been compiled to provide radar practitioners with a unique reference source covering the latest developments in micro-Doppler analysis, extraction and mitigation techniques. The article shows that this research area is highly active and fast moving and demonstrates that micro-Doppler techniques can provide important solutions to many radar target classification challenges.

/pdf/developments-in-target-micro-doppler-signatures-analysis-23facvvnxd.pdf

Developments in target micro-Doppler signatures analysis: radar imaging, ultrasound and through-the-wall radar

In this report, two novel algorithms for ground moving target classification using additional information features related to the radial velocity variability are proposed and studied. These approaches are compared with feature extraction method based on linear prediction model and cepstrum coefficients. Results of computer simulations performed by exploiting real-life radar backscattering signal records with different decision making concepts demonstrate the benefits of suggested approach.

Classification of ground moving radar targets by using joint time-frequency analysis

It was shown that the methods of nonlinear dynamics, surpassing traditional methods of temporal, frequency or frequency-temporal analysis of dangerous ignition factors may be used for early detection of ignitions in premises. The existence of carbon monoxide in gas medium was found to be most dangerous at fires in premises. The theoretical grounds for studying recurrent diagrams of carbon monoxide concentration in gas medium were substantiated. The modification of recurrent distance diagrams, based on power representations was proposed, making it possible to highlight selectively or to smooth structural features of configuration of recurrent points of distance diagrams. Results of research into recurrent diagrams of dynamics of carbon monoxide concentration show that the specified factor of ignition of materials has generally not stochastic, but chaotic dynamics. It was qualitatively determined that dynamics of carbon monoxide concentration in gas medium has non-uniformed distribution of points. In this case, the configuration of clustering of recurrent points of diagrams for various flammable materials varies and can be used for detecting the type and the beginning of early ignition of combustible material. The established fact of chaotic dynamics of carbon monoxide concentration in gas medium at early ignition of materials should be taken into consideration in the development of new technologies for reliable detection of early ignitions in premises. The data, obtained in the research, are important for deeper understanding of dynamics of the process of carbon monoxide formation in gas medium in non-airtight premises at ignition of various materials, because it is related to saving lives of people who are in these premises and their timely evacuation.

/pdf/studying-the-recurrent-diagrams-of-carbon-monoxide-4j5sys7g1p.pdf

Studying the recurrent diagrams of carbon monoxide concentration at early ignitions in premises

The concept of guaranteed ignition detection at a site was introduced. A criterion of optimization of guaranteed detection was formulated, which comes down to the equality of probabilities of false detection and missing of ignition source. Algorithms and structure of fire detectors, capable of self-adjusting by ignition of materials, were developed. Their distinctive feature is the possibility of being applied under uncertain conditions for arbitrary and combustible materials that are unknown in advance. To enhance effectiveness of fire detectors capable of self-adjusting by combustion of materials, we proposed adaptation of original threshold value to current observations of ignition components. For this purpose, it was proposed to use the procedure of median filtration of recorded data. As a parameter of convergence of the procedure of threshold self-adjustment, we consider a fixed and dynamic way of its determining. This makes it possible to provide adjustment of original convergence of procedures toward observed components of combustion of various materials. Verification of the proposed self-adjusting fire detectors indicates their capability to provide guaranteed detection of sources of ignition for various materials at the early stages under conditions unknown in advance

/pdf/design-of-fire-detectors-capable-of-self-adjusting-by-2kqm9p9osb.pdf

Design of fire detectors capable of self-adjusting by ignition

A general analysis of self-adjustment algorithm of fire detectors for early guaranteed ignition detection was carried out. It was shown that step magnitude in the algorithm of self-adjustment of fire detectors can be a fixed or a selected variable, depending on the level of registered actual data about the factor of ignition of combustible material. Specific features of fire detectors, self-adjusting by ignition, relate to a non-linear nature of the algorithm and conditions for providing guaranteed ignition detection. In this case, for effective self-adjustment of fire detectors, it is expedient to set the initial threshold value by registered actual data in the absence of ignition and to regulate the adaptation step size in a special way. It was shown that the fact of an increase in self-adjusted threshold relative to its original value can be a sign of ignition detection. For probabilistic assessment of the fact of ignition detection, it was proposed to use exponential smoothing of characteristic function, which allows generation of dynamic assessment of probability of ignition detection. A study of dynamics of self-adjusting threshold and probability of ignition detection indicate capability of self-adjusting fire detectors to provide early guaranteed detection of different sources of ignition of flammable materials under conditions unknown in advance

/pdf/research-into-dynamics-of-setting-the-threshold-and-a-3rmu7j8qxq.pdf

Research into dynamics of setting the threshold and a probability of ignition detection by self­adjusting fire detectors

We have considered the application of the method of nonlinear dynamic systems in order to analyze and detect the structural patterns in the dynamics of increments in the state of a gas medium generated by early ignitions of materials in a non-sealed chamber. The research method is based on analysis of the correlation dimensionality of increments in the state of a gas medium during ignition of materials. We have theoretically justified the method for evaluating the dynamics of correlation dimensionality of increments in the state of a gas medium at ignition. The considered method for CD evaluation is based on the computation of the Grassberger-Procaccia correlation integral, applied to the gas medium state increments using a sliding window with a fixed width. That allowed us to derive a current estimate of CD increments in the state of the gas medium during ignition of flammable materials in a chamber synchronized with the observation data acquisition rate. We have analyzed the dynamics of correlation dimensionality of increments in the state of a gas medium at early ignition of alcohol, paper, wood, and textiles in a simulation chamber. It was established that for the investigated state of the gas medium during ignition of various examined materials, the dynamics of correlation dimensionality is within 0.1 to 0.6. It is noted that this fact testifies to the fractal structure of the considered increments in the state of a gas medium in a chamber and its chaotic dynamics at the emergence of ignition sites of tested materials. In this case, the fractal structure is not the same, suggesting a "transitional chaos" in the examined state of the gas medium. It was established that current estimates of the correlation dimensionality of increments in the state at the time of materials ignition tend to a sharp increase. A given fact can be used to reliably detect early fires indoors. The results obtained are important for the in-depth studying and understanding of patterns in the structure of dynamics of increments in the state of a gas medium at early ignition. It has been shown the increments in the states of a gas medium at premises characterize it as a chaotic dynamic system with a small fractal dimensionality as opposed to the traditional approach assuming a gas medium being either deterministic or random system

/pdf/analysis-of-correlation-dimensionality-of-the-state-of-a-gas-9lz5em2svn.pdf

Analysis of correlation dimensionality of the state of a gas medium at early ignition of materials

We presented results of experimental research into dynamics of hazardous factors of the state of environment, as well as of their autocorrelations and pair correlations in a simulation chamber at various lags for early detection of ignition of combustible materials, such as alcohol, paper, timber and textiles. The study was carried out based of Pearson coefficient of lag correlation that possesses the best interpretability of results. It was found that dynamics of factors of the state of environment in space of the studied factors prior to ignition of materials is locally inhomogeneous and distributed with random transitions between local states. At ignition, in the studied environment, there occurs grouping of its distributed local states depending on material of ignition. It was established that the character of grouping can be used as an indicator of early detection of ignition and identification of the type of combustible material. In this case, early ignition causes a slight increase in constant of time of fluctuations of the studied factors of environment. However, this indicator has insufficient differentiability for early detection of ignitions. A more informative indicator is a degree of mutual correlation of fluctuations of environmental factors at zero lag. It was found that for alcohol, paper, timber and textiles, an effective indicator of early detection of their ignition is coefficient of pair correlation of CO and temperature, the module of which is close to 1, and for textiles – to –1

/pdf/results-of-experimental-research-into-correlations-between-2d5kh3ta0b.pdf

Boris Pospelov

Papers

Studying the recurrent diagrams of carbon monoxide concentration at early ignitions in premises

Design of fire detectors capable of self-adjusting by ignition

Research into dynamics of setting the threshold and a probability of ignition detection by selfadjusting fire detectors

Analysis of correlation dimensionality of the state of a gas medium at early ignition of materials

Results of experimental research into correlations between hazardous factors of ignition of materials in premises

Boris Pospelov

Papers

Studying the recurrent diagrams of carbon monoxide concentration at early ignitions in premises

Design of fire detectors capable of self-adjusting by ignition

Research into dynamics of setting the threshold and a probability of ignition detection by self­adjusting fire detectors

Analysis of correlation dimensionality of the state of a gas medium at early ignition of materials

Results of experimental research into correlations between hazardous factors of ignition of materials in premises

Research into dynamics of setting the threshold and a probability of ignition detection by selfadjusting fire detectors