ARMA model orders and parameters estimation using evolutionary algorithm

العناوين الأخرى

تقدير درجة و معاملات متسلسلات المتوسط المتحرك التدريجي الذاتي باستخدام الخوارزميات المتطورة

مقدم أطروحة جامعية

al-Qawasimi, Khalid Isa

مشرف أطروحة جامعية

al-Hamami, Ala Husayn
al-Smadi, Adnan Muhammad Khayr

أعضاء اللجنة

Ubayd, Nadim Ali Miri
al-Ani, Muzhir Shaban
Qaṭawinah, Muhammad Sulayman

الجامعة

جامعة عمان العربية

الكلية

كلية العلوم الحاسوبية و المعلوماتية

القسم الأكاديمي

قسم علم الحاسوب

دولة الجامعة

الأردن

الدرجة العلمية

دكتوراه

تاريخ الدرجة العلمية

2010

الملخص الإنجليزي

Autoregressive Moving Average (ARMA) model is an active research area that has many major roles in a variety of application fields.

For examples, ARMA process plays an important role in modeling the Internet traffic of a network, financial market forecasting, biomedical signals analysis, speech modeling, radar, sonar, linear prediction, system identification, and spectral analysis.

Model order selection of a general ARMA process has been of considerable interest for some time and it has a long and continuity history.

ARMA model is one of the most important time series models.

An important step in system identification and time series prediction is the estimation of the model order.

The fundamental problem of the ARMA modeling is that, the model order is not known and it needs to be estimated before solving the parameters estimation problem.

Hence, determining the order of the ARMA model is the first step to estimate the model parameters.

Selecting the optimal ARMA model order is difficult and has never been obtained satisfactorily.

This dissertation provides two novel algorithms for solving the problem of ARMA model order estimation.

The two proposed algorithms are based on the algorithm proposed by Liang et al.

The observed sequence may be contaminated by additive Gaussian noise.

Simulation examples are given to demonstrate the performance of the proposed algorithms at different Signal-to-Noise-Ratios (SNRs).

The computations were performed in MATLAB.

The first proposed algorithm, called Pivot-Neighbors comparisons (PNC), is based on a rounding approach which uses the floor and the ceiling functions.

The rounding approach is implemented to deal with the precision of binary words.

The proposed algorithm is based on selecting a sequence of pivot cells from the Minimum Eigenvalue (MEV) of a covariance matrix computed from the observed data.

It searches for a corner that contains the estimates of the true orders using the floor and the ceiling functions of the pivot cells values and the values of its neighbors.

The second proposed algorithm is based on modeling and designing Artificial Neural Network (ANN) architecture for a special matrix constructed from the MEV criterion to estimate the ARMA model orders.

The proposed ANN-based algorithm is based on training the MEV dataset using the Back-Propagation (BP) learning algorithm.

Our goal is to develop an automated system; hence, the model order can be selected automatically without the need of prior knowledge about the model or any human expert participation.

The empirical results of several ARMA model examples indicate that the proposed algorithms have improved estimation accuracy over the MEV method

التخصصات الرئيسية

الرياضيات

عدد الصفحات

192

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-528454