IRAQI JOURNAL OF STATISTICAL SCIENCES

Abstract
This paper deals with a new proposed method to measure the activity of internet sites linked to a certain network. This method is then applied on the Web sites associated with the University of Mosul. The proposed method depends mainly on the principal components technique and from which is extracted the most effect factors on the activity of Web sites. It transforms the principal components to a classification equation taking into account the importance of each influential principal component. It has been used to classify websites of colleges associated with the University of Mosul, according to the proposed method, and it gives encouraging results.

Abstract

In this paper, a new nonlinear conjugate gradient methods based on the modified secant condition is derived which are given by Li and Fukushima (Li and Fukushima, 2001). These methods showed global convergent under some assumptions. Numerical results indicate the efficiency of these methods to solve the given test problems.

ABSTRACT
This research includes the application of some statistical techniques for studying the time series of the average monthly rainfall as an output series with two of the variables which affect on to, which are the series of the average monthly relative temperature and humidity as an input which is measured at the meteorological station of Ninavah. The techniques used are the modeling by an(ARIMA) model as well as the dynamic regression model. So that the perfect dynamic regression model selected was suitable for determining the future forecasting values.

Abstract
Regression splines is one of the methods that are used to estimate the regression curve non parametrically. One of the most important elements that contribute to the application of the method is to determine the degree of the spline function and the number of knots and their locations Choosing the number of knots and their locations is the main problem in estimating the non parametric regression using regression splines and, when carefully selected, the amount of smoothing in the fitted curve will be in optimal conditions. The research is to shed light on two methods of knots locations; the first method includes place of the knots which represents the data quintile and use one of models selection criteria such as ( Generalized Cross Validation) criterion to select the number of these knots. The other method depends on placing the knots on equal spaces and the use of criterion to select the final locations of knots and its number , A comparison is made between the two methods using one of the errors criteria which is ( Mean Average Absolute Error),depending on the experimental data using the simulation method . Through simulation experiments the method of place knots in equal spaces , is better than the method of placing knots in the form of quintile of data .

Abstract
In this paper an image is hidden in another image by applying logical XOR operation to the two least significant bit of each byte of the cover and two bits of the secret message to produce the stego-cover image which is used as an input with the cover to Radial Basis Function Network (RBFN) or General Regression Neural Network (GRNN) to produce the weights.
Cover is delivered once to the recipient who can use it for unlimited number of messages. The weights are delivered to the recipient for each hidden message as a key. The recipient uses the cover with the weights to unhide the message, so that this method includes two levels of security. The first one is hiding the message in the cover to produce stego-cover image. The second one is ciphering the embedded image using RBF Neural Network or GRNN. The stego-cover is considered as a target and the input to the neural network is the cover image. Then the weights, which represent the encrypted information are reconstructed by training the neural network. The recipient can use RBF Network or GRNN to unhide the message by having the stego-cover image then the message.
In this paper the performance of RBF Neural Network and GRNN are compared. Noted that the GRNN is better than RBF for this work. Matlab R2008a was used in this paper.

Abstract
This paper deals with the two-stage hybrid flow shop problem , in which the first stage consists of three machines , the second stage consists of two machines . The aim is to find out the optimal scheduling for n jobs when processing in this environment when the makespan is minimum. Therefore we propose a particle swarm algorithm which consists of a new procedure to calculate the makespan and a new stopping criteria .Also, we added improvement to the proposed algorithm , by using one of the components of the genetic algorithm (crossover operation) in order to obtain initial swarm particles instead of random obtaining. After applying the two algorithms on several problems which were generated randomly by uniform distribution , the results showed that the improved proposed particle swarm algorithm was the best in finding out the optimal scheduling for jobs and in cpu time to reach the solution.

Abstract:
In this paper we determine the optimal Bayesian sample size to estimate location parameter of normal distribution when the shape parameter is unknown under squared loss function and linear exponential loss function (LIENX) .In application we explain the efficient estimator under two loss functions and obtain the relationship between the sampling cost per unit and the sample size which we determine using Matlab language.

Abstract
This research deals with the most important methods of estimating parameters of the Geometric Process(GP). We use the least squares and the modified moment method, as nonparametric methods, and the maximum likelihood method as a parametric method with the lognormal distribution as the probability distribution for the first stop. To compare these methods we use a criteria proposed in this research, as well as comparing(GP) with the renewal process(RP). As a part of applied research we test whether the data which represent the occurring stops in

Abstract:
The research deal with the α-series stochastic process. It’s proposed by the researchers [Braun et a., 2005] as considered a very useful complement to the increasing geometric stochastic process. The parameters of α-series process are estimated using some nonparametric methods, as well as Mann test for trend analysis monotonous stochastic process, and technical graphic that tests the
ownership of the process to the α-series process. The research includes an application addresses the stops of Cement production factories in Iraq. The statistical analysis is performed for the factory stops as α-series process, and tests the appropriateness of the data for the process. The study also estimates the expected number of factory stops cases, and compares between the suggested estimation methods.

Abstract :
The idea of this research is based on adopting Support Vector Machine for Regression (SVR) in estimating and predicting regression models and comparing it with one of the multi layer neural network that has back propagation((MLP). Therefore ,this will lead to demonstrate the advantages and possibility to use it in scientific application by illustrating how it is accurate in comparison with the ((MLP) by using Mean Square Errors(MSe) .
The time sequence data are that used for month flow for the water from Tigris that incomes in to Mosul city for the period between (1950-1995) to accomplish the comparison process by using the employment of two computers softwares.

Abstract
In this paper GRNN is used to hide an image into a cover file, Audio and video files are used as cover files in this work. The hiding was applied using one of concealment algorithm that is (Least Significant Bit (LSB)) by applying (XOR) operation between the last (2bits) of each byte of the image (cover) and 2bits of the message to produce the
(stego_cover), which feed with the cover to General Regression Neural Network (GRNN) to produce the weights.
Cover is delivered once to the recipient who can use it for unlimited number of messages. The weights are delivered to the recipient for each hidden message as a key. The recipient uses the cover with the weights to unhide the message. So, this method includes two levels of security. The first one is hiding the message in the cover to produce embedded cover (stego-cover). The second one is ciphering the embedded cover using GRNN Neural Network. This Network is considered as a target and the input to the Neural Network is the cover. Then the weights, which represent the encrypted information, are reconstructed. The recipient can use GRNN Network to unhide the message by having the embedded cover (stego-cover) then the message. Matlab R2008a was used in this paper.