Mustansiriyah journal of pure and applied sciences

The mathematical models suggested in this article accounts for two types of predator-prey
system. It was suggested that the model includes SIS (susceptible infected susceptible) disease
in prey types, which is expansion by proximity between infected and susceptible species, and
there is harvest in predator species.The epidemic cannot be transmitted from prey to predator
during the predation process. The equations that represent the interaction between infected
individuals in the prey population and susceptible individuals in the predator population are
ordinary differential equations. Based on what was assumed above, any Possible balance points
can be analyzed using methods of mathematical .Depending on the numerical simulations, the
dynamics of the system’s behavior were studied local and the dynamic behavior was studied
globally using Lyabnov functions, and the variation of harvest and disease on dynamics of
model’s behavior was also discussed.

In this paper, non-linear fuzzy Volterra integral equation of the second kind (NFVIEK2) is
considered. The modified decomposition method will be used to solve it. Some numerical
examples are prepared to show the efficiency and simplicity of the method.

Signature verification can be consider as one of the most popular ways for verifying the
authenticity of legal contracts and documents, it is also used to verify individual signature on
banking and several financial application. There are numerous methods for signature
verification, starting from the old traditional manual approaches to the ultramodern techniques.
In this paper, the tools available through cloud computing were employed to develop a
signature verification platform in banking transactions quickly and easily in a way that achieves
higher accuracy and avoids the errors that may accompany previous methods. The use of cloud
computing has become essential in all fields due to its availability at all times and in all places,
in addition to the ease, speed and accuracy provided by cloud services. The proposed platform
matches the client’s signature uploaded from one of the bank branches with a database
containing the signatures of all clients in all branches of the bank to verify the validity of the
submitted cheque. The final results of the proposed platform show the high accuracy, reliability
and fast performance.

This study investigates the use of advanced machine learning algorithms to classify different
classes of epilepsy EEG signals. Using the Time Series Feature Extraction on basis of Scalable
Hypothesis tests (TSFRESH) package, a large set of features was reduced to the most important
ones by extracting them from the University of Bonn EEG dataset. Following that, seven
distinct machine learning classifiers were trained using these features. Three different
classification tasks are used to evaluate the classifiers using cross validation. The evaluation
results of each algorithm after testing them on each of the three tasks show that the accuracy of
the Support Vector Machine (SVM) reaches 99.60% for binary classification, the accuracy of
Histogram-based Gradient Boosting Trees (HGBT) reaches 99.20% for the three-class task, and
for the five-class classification, the Random Forest achieves 94.80% accuracy, which is the
highest among the others. These results show that there is no dominant algorithm that works for
all classification tasks and it is necessary to always run more than one algorithm to get a better
assessment.

The need to build visual speech recognition is the main motive that made us research this topic.
There is difficulty in understanding speech when visible, there is an urgent need to develop a
system that can read lips and understand visual speech. This would help people with hearing
impairments to interpret lip movements and understand the spoken phrases and sentences. It
would also assist people in noisy environments such as stadiums, airports, factories, and other
places where it is difficult to access audio signals. Therefore, researchers are continuously
striving to find the best solutions to address this problem. This system is of great importance to
empower people with hearing disabilities and others who face difficulties in understanding
spoken language in noisy environments.
In an attempt to solve this problem, this research designed and implemented a real-time
system to visually interpret and understand spoken phrases and sentences without the need for
audio input. The proposed system consists of two main stages: the first stage is detecting the face
region and the mouth region, followed by the detection and localization of the area of interest
(ROI), which is the lip region. The process is carried out by capturing video of the speaker and
dividing it into consecutive frames. After detecting the face region and the mouth region, the
area of interest, which is the lip region, is detected. Multitask convolutional neural networks
algorithm (MTCNN) was used to perform the detection and localization of these regions. The
second stage involves inputting the frames corresponding to the lip region into The squeeze Net
convolutional network model for recognizing the spoken phrases and sentences. The proposed
method achieved an accuracy of 90%.

In this research we provided new concept namely supra–semi extremally disconnected space
(briedly SU-SED spaces) by using semi open set via supra topological spaces and we study the
properties of this space, we defend another types of sets such as supra semi closure and supra
semi interior on supra semi extremally disconnected space. Like that we knew another supra
topological spaces is called supra- semi hyper connected space shortly by SU- SH space we
proof supra Semi hyper connected spaces is strong from of supra- semi extremally disconnected
space that mean every SU-SH space is SU-SED but the converse not true and we explained an
example indicates the validity of the result. We study relationship between supra–semi
extremally disconnected space and supra-semi hyper connected space, some facts on this space
and results have been given, we expanded to study the product of two spur semi extremally
disconnected.. Finally we give example to support our work.

Several areas of AI have found uses for face recognition (FR) technologies, such as biometrics,
authentication, security, surveillance, and law enforcement. Results from FR using deep
learning (DL) models, especially CNNs, have been promising. By utilizing learning methods
such as Convolutional Neural Networks (CNN) and mesh algorithm to improve face
recognition accuracy. our goal is to ensure privacy protection while tackling issues related to
varying lighting conditions and facial expressions. The proposed model (CNN-Mesh) is
combined CNN algorithm with mesh technique for prediction the fully landmark of face. the
CNN-Mesh approach performs in comparison to that of other facial recognition methods that
are currently in use. The purpose of this study is to determine whether or not the integrated
model offers improved resilience against variables such as varied facial expressions, age, or
cosmetics, in addition to speed and computing efficiency. The purpose of this research is to
determine the extent to which merging CNNs with Mesh algorithms results in improvements in
both accuracy and processing speed. The combination has two stage: first stage is that the
network collects feature extraction to solve the classification problem. Second stage apply
mesh network after collects whole features for used it to describe the edge of faces. This study
investigates how convolutional neural networks (CNNs), which are particularly effective at
extracting hierarchical information from images, with used to provide Mesh networks, which
are primarily concerned with predicting certain face landmarks, with a more accurate input.
The purpose of this study is to determine which elements retrieved by CNNs (such as edges,
textures, and deep semantic representations) are the most helpful for the succeeding Mesh
network to reliably predict face structures. Moreover, it able the network to construct
predictions based on those traits. During the process of learning, the model is instructed to
recognize patterns and traits in the images that are associated with certain individuals. Our
strategy involves the use of 68 Landmark markers to optimize facial recognition precision
resulting in advancements in this field. The results of our CNN-Mesh algorithm achieved a 97%
accuracy rate for classification and detecting faces, while the accuracy of the model is increased
after mesh network to 99%. Moreover, we train ResNet model for comparative with our model
which achieve accuracy 31.73%. while CNN 88.46%.

A hybrid solar photovoltaic system can supply homes and all government and private sector
buildings with electricity in cities and remote places. Solar PV systems generate emission-free
electricity that is stable, safe, noise-free, easy to use and does not require refuelling. It also
reduces the use of fossil fuels in power plants and reduces pollutants and greenhouse gas
emissions that cause climate damage, Therefore, the current research aims to design and
implement a hybrid photovoltaic solar system to provide one of the laboratories of the
Department of Physics at the college of education, Mustansiriya University, with electrical
energy during a year. The total electrical energy consumed by the loads in the laboratory
(devices, lighting lamps and fans) was measured by about 2000 watts. The working hours of
the laboratory were estimated at six hours. In light of that, a photovoltaic solar system
consisting of eight solar panels with a capacity of 545 watts per panel was designed in standard
conditions and a total capacity of 4360 watts. After the design and installation, the laboratory
was operated at its total capacity and the power supplied by the panels to the laboratory was
recorded during the first daylight until after midday. The results were encouraging, as it was
found that the net power processed by the system is about 3500 watts. Therefore, we can
conclude that the construction of solar systems is an economically feasible process under the
conditions of severe shortage in the supply of electrical energy and the high cost of purchasing
electrical energy from the private sector.

With the increasing use of fake images and videos on social media platforms, the issue of
verifying the authenticity of images has become of great importance for privacy protection
.Recently, deepfake images technology has widely used for face-swapping in order to generate
fake or forged data to deceive society. Detecting the rightfulness of images has become
progressively critical issue due to the potential harmful effect on the human privacy and
security. The main objective of this study is to develop and implement a deep fake detection
system using face region extraction method and YOLOv8 deep learning model. The proposed
system mainly relies on extracting the face region from the input images sample using Multi
Task Cascaded Convolutional Neural Networks (MTCNN) method to reduce the processing
time of fake /real face image in the detection process. The extracted faces regions are then
analysed and classified based on YOLOv8 model to obtain the fake and real information
related to each input image in term of binary classification. Further, the main hyperparameters
of Yolov8 model were tuned throughout model training phase to generate more robust trained
model for achieving higher detection accuracy. The system performance was evaluated using
multiple fake and real images datasets. The results showed that the proposed system achieved a
detection accuracy = 97.5%, Precision = 97.03%, Recall = 98%, and mAp = 99%

The operation of an optical neural network via feed-forward (FF) configuration is
experimentally simulated in the laboratory. The FF setup is tested using optical injection (OI),
and the behavior of follower laser diodes (FLDs) subjected to chaotic modulation is examined.
The last two laser diodes (LDs) are exposed to different weights of chaotic modulated signals
through optical filtration and attenuation. Observations of the emissions from these two FLDs
during FF operation are verified by frequency spectra calculated from time series data. Signal
broadening is assessed by measuring the full width at half maximum (FWHM), and chaotic
signal spikes are analyzed by counting the number of peaks associated with signal amplitudes
for the FLDs. Additionally, LD control parameters, including the bias voltage of the influencer
laser diodes (ILD1, ILD2) and two additional FLDs, are examined. A maximum FWHM of
1.25 GHz for FLD2 is observed with a bias voltage of 3.6V and a modulated signal attenuation
of -12dB. To determine the synchronization state, the correlation between the ILDs and FLDs
is calculated. Results indicate fluctuations between negative and positive values, with the best
correlation value being 0.4. These results confirm anti-synchronized ILD-FLDs, which is
crucial for ensuring privacy in transmitting units within a chaotic optical communication
system simulating an optical neural network.

Abstract: An age estimation term refers to the procedure of expecting a human’s consecutive
age based on different factors or indicators. These factors could be physical features like the
color of the hair, wrinkles, and the shape of the body. Also, could be a biological feature like
DNA methylation or dental growth. The approaches or techniques used in age estimation could
be as simple as visual opinion or more complex as analyzing diverse databases using machine
learning methods and algorithms. The algorithms used features extracted from images captured
using various technologies, one of these technologies is the Kinect sensor due to its ability to
provide facial, skeleton, and body features. In this study, age estimation applications, human
age features, database, approaches used for age estimation, Kinect sensor abilities, limitations,
and improvements algorithms, are presented with the most related works, the final section is for
the conclusions.

There are limited published data regarding the recent incidence trends of patients with
malignant neoplasms in ninewah governorate. This study aims to analyze the time series
properties using the (BOX & Jenkins) approach in the analysis (estimation, identification, and
selecting the appropriateness of model for prediction). In this study the capabilities of A time
series are employed to determine the best and most efficient statistical model for the purpose of
prediction the number of people with this disease. The findings of the data analysis indicated
that ARIMA (2,1,0) is the most suitable model for predicting the quantity of people with
malignant neoplasms in Ninewah Governorate depending on the data for the period of (2016 -
2021). The results explained the type of this function are an nonstationary series on average,
and there is a clear general trend in the series. The stationarity of the time series was achieved
after taking the first difference of the data, and after matching the auto and partial correlation
coefficients of the time series with the theoretical behavior of the auto and partial correlation.
The number of patients with malignant neoplasms was foretold using this model.

Land use/ land cover (LULC) changes over time and is a crucial indicator of understanding any
region environmental and economic dynamics. Satellite remote sensing data provide a valuable
source of information to monitor LULC changes at different spatial and temporal scales. This
study used a time series of satellite image outputs to analyze LULC changes in Babylon
Governorate over a period extended from 1992 to 2020. The study methodology involved
processing and analyzing a time series of classified digital satellite images, calculating the
areas of the classes and plotting the time series curves of 13 LULC classes within the study
area, based on data prepared by ESRI. The data were processed using ArcGIS Pro. V.3.3,
represented by spatiotemporal distribution maps of LULC, and time series curves to understand
the driving forces behind these changes. The results identify the changing trend of each class in
the area, in an increasing or decreasing manner. The results also, show significant changes in
the study area, including a substantial rise and expansion of the urban regions, and fluctuations
in vegetation cover and barren lands. A noticeable decrease in the areas of rainfed croplands is
recorded. This is due to the rainfall decreasing during the study period, as the region is located
in soils characterized by chemical degradation, i.e., increased soil salinity. The information
derived from this study can help develop sustainable land management strategies, support
urban and regional planning efforts, and contribute to understanding the complex interactions
between human activities and the natural environment.

Spirogyra varians were taken from 2nd February to 21st April 2023 from the water channel in
Diyala Bridge and the lake at the University of Baghdad, Various parameters including
temperature, acidity, turbidity, electricity, and the presence of nutrients nitrogen (NO3) and
phosphorus (PO3) were recorded. Following the identification of Spirogyra varians, an algal
extraction was conducted using methanol as the solvent. The algal extract was utilized to
synthesize MgO nanoparticles (MgONPs), 0.05 M concentrations of Mg (NO3)2H2O with
reaction degree equal to 80º at 6 pH were used as conditions of the green synthesized. The green
generated MgONPs were characterized using Ultraviolet – Visible spectrophotometry, Field
Emission Scanning Electron Microscope (FESEM), energy-dispersive X-ray spectroscopy
(EDS), and X-ray diffraction (XRD) techniques. The anticancer efficacy against MCF-7 breast
cancer cells and its correlation with Reactive Oxygen Species (ROS) were assessed. The
findings demonstrated significant efficacy and impact of the produced magnesium oxide
nanoparticles (MgONPs) against MCF-7 breast cancer cells. Additionally, a clear correlation
was seen between MgONPs and reactive oxygen species (ROS) in the cancer cell line.

Extremist groups often utilize social media to disseminate their ideas and beliefs in order
to compile more new members and thereby help them spread violent content and
extremist ideologies that threaten social cohesion. The goal of detecting extremism on
social media is to identify and prevent the spread of these ideas through the use of
artificial intelligence technologies that will help us detect extremist texts in social media.
This paper proposes a methodological approach based on the use of deep learning
algorithms for the effective cessation of extremist texts spread. The dataset we used is
called ISIS radical annotated tweets; it consists of 24078 tweets from 174 accounts
related to the extremist organization known as the Islamic State. Initial preparation and
cleaning of the dataset are applied to ensure the accuracy of the dataset. Feature
extraction techniques using (Term Frequency (TF), Inverse Document Frequency (IDF),
Word2vec) are applied to extract most important features. To identify extremist texts,
this paper chooses two state-of-the-art deep learning algorithms Long Short-Term
Memory (LSTM) and combined Convolutional Neural Network CNN- LSTM.LSTMs
have a lengthier term of memory and thus work well with text content, whereas CNNs
are designed to extract features from the data. The final CNN-LSTM model combines
both algorithms’ strengths. In this study we used Deep learning techniques to classify
extremist texts, two different deep learning approaches were used (LSTM and Hybrid
CNN-LSTM). The CNN-LSTM combination produced the highest accuracy reached to
(98.20%).

This paper introduced a new type of fuzzy soft quasi paranormal operator (QFSPO) in Hilbert
space (HS) and examined some of its main properties. To improve the operator\'s efficacy, it
was compared to other operators in certain conditions. The fact that this type is interesting has
been supported by very important theories for example, the possibility of applying the proposed
operator condition to the inverse if it exists, and other theories.

The search for alternative antimicrobial agents has continued to be a top priority due to the
growing cases of resistant strains of bacterial and fungal species. This work evaluated the
antimicrobial activity of some known mandelonitrile derivatives. Thus, the compounds were
successfully obtained by reacting the appropriate aldehydes with the solution of sodium
metabisulphite and sodium cyanide at room temperature. The structures of the compounds were
confirmed by the NMR and FTIR techniques. Evaluation of the susceptibility of the compounds
against four bacterial and one fungal species using disc diff method showed excellent zones of
inhibitions for compounds 2b, 2d, 2e, and 2f against Pseudomonas aeruginosa. (30.1 mm – 17.2
mm) compared to the control drug, piperacillin (25.8 mm – 13.4 mm). The minimum inhibitory
concentration (MIC) and minimum inhibitory concentration (MBC) of compounds 2b, 2d, 2e,
and 2f were similarly found to be lower against the same organism. All the synthesised
compounds demonstrated mild activity against the fungal specie, E. coli. This work has
demonstrated the potential of the studied nitrile-containing compounds against the investigated
bacterial and fungal species.

As cloud computing becomes increasingly integral and the deployment of Internet of Things
(IoT) devices accelerates, safeguarding the integrity and confidentiality of transmitted data has
emerged as a paramount concern. This research introduces a comprehensive security
framework that leverages advanced cryptographic protocols—specifically, the integration of
Pretty Good Privacy (PGP) concepts with robust encryption algorithms Advanced Encryption
Standard in Galois/Counter Mode (AES-GCM) and RSA (Rivest–Shamir–Adleman) for secure
key distribution. In this architecture, AES-GCM ensures both the confidentiality and integrity
of IoT messages, while RSA encryption is employed to securely transmit the session keys
among users. Once decrypted, the IoT data undergoes classification through a Long Short-Term
Memory (LSTM) neural network, whose hyperparameters are fine-tuned using the Snake
Optimizer algorithm (SO) to enhance detection performance. The framework’s effectiveness is
evaluated using the TON IoT dataset, which encompasses realistic attack scenarios.
Experimental findings indicate a superior classification accuracy of 98%, demonstrating the
system’s ability to reliably differentiate between various attack types and normal network
behavior. These results underscore the critical role of combining strong encryption
methodologies with state-of-the-art machine learning techniques to advance security within
cloud-based IoT environments.