Search

One of the chronic diseases is Diabetes, Diabetes is a metabolic disorder category caused by continued high levels of blood sugar. It is regarded as one of the most deadly diseases in the world. If accurate early prediction is possible, Diabetes severity and risk factor can be significantly lowered. Machine learning has become more popular in the medical community as a result of its ascent, and in the field of diseases in particular. In this paper, we propose a model that can predict where the patient has or hasn't have diabetes. Our model is based on the prediction precision of certain powerful machine learning (ML) algorithms based on different measures such as precision, recall, and F1-measure. The Pima Indian Diabetes (PIDD) dataset has been used, that can predict diabetic onset based on diagnostics manner. The results we obtained using Logistic Regression (LR), Naive Bayes (NB), and K-nearest Neighbor (KNN) algorithms were 94%, 79%, and 69% respectively. The results show that LR is more efficient at predicting diabetes Compared to other algorithms.

Humans are emotional beings. When we express about emotions, we frequently use several modalities, whether we want to so overtly (i.e., Speech, facial expressions,..) or implicitly (i.e., body language, text,..). Emotion recognition has lately piqued the interest of many researchers, and various techniques have been studied. A review on emotion recognition is given in this article. The survey seeks single and multiple source of data or information channels that may be utilized to identify emotions and includes a literature analysis on current studies published to each information channel, as well as the techniques employed and the findings obtained. Ultimately, some of the present emotion recognition problems and future work recommendations have been mentioned.

Optical network is the corner stone for the high data rate/throughput applications due to the wide bandwidth allowed in optical channel. Additionally, optical network tackled the delay problem since data is reformed into equivalent light and transmitted light speed to the destinations e.g. 3 × 10 8 m/h. Data can travel in optical domain over the light-path in high rate approaching to tens of Gigabits. Routing of data in optical domain is controlled by the wavelength division multiplexing (WDM) scheme in which allotting each wavelength at the ingress of optical fiber with its suitable bandwidth. Bandwidth of optical path is being optimized in order to utilize the fiber resources efficiently using Artificial Neural Network (ANN). In this paper Artificial Bee Colony (ABC) algorithm is used for optimizing the performance of ANN which is predicting the wavelength of the optical path.

Levenshtein is a Minimum Edit Distance method; it is usually used in spell checking applications for generatingcandidates. The method computes the number of the required edit operations to transform one string to another and it canrecognize three types of edit operations: deletion, insertion, and substitution of one letter. Damerau modified the Levenshteinmethod to consider another type of edit operations, the transposition of two adjacent letters, in addition to theconsidered three types. However, the modification suffers from the time complexity which was added to the original quadratictime complexity of the original method. In this paper, we proposed a modification for the original Levenshtein toconsider the same four types using very small number of matching operations which resulted in a shorter execution timeand a similarity measure is also achieved to exploit the resulted distance from any Edit Distance method for finding the amountof similarity between two given strings.

Face recognition systems played a vital role in the last decades, artificial intelligence techniques are exploited to perform such systems. Time and Accuracy of real-time systems are very important evaluation criteria. This paper proposes a real-time face recognition system using software agent, three world datasets are experimented as static images, 98% recognition rate is achieved, and 90% for real-time face recognition, with one frame per second, which is a significant percentage in comparison with related works.

In this Paper an Approach For brain tumour types detection and classification is proposed. This study was developed using agency system classification brain tumors. It includes sensor to get the input image. The agency program comprises preprocessing to reduce the salt and pepper noise. The detection of tumor is materialized by using multi-level threshold segmentation, opening and closing morphological operation to enhance the multi-level threshold, watershed transform to segment the tumor, extract important feature using binary object feature, and classification the types of brain tumor. The actuator is used to output the classification and decision. The agency system achieves an accuracy (100%) to classify the images into normal tissues, meningioma tumor, meningioma benign, Gliomas and metastases.

Nowadays intelligent transportation system has become an active research field, automated vehicle identification is the primary aspect for achieving good intelligent traffic analysis, this paper illustrates a comparative study among three techniques that used for identifying vehicles: ANPR system, Barcode system, and RFID system. In addition, this paper describes similarities, differences, and challenges.

Video compression techniques play a vital role in storage & transmission through a limited bandwidth. As a resulted containing video files repeated sequence of still image, each second represented by 24 images or more which is represent the motion in video file. The repeated sequence generates a large amount from temporal redundancy through similarities of sequence image with little difference between them, in addition to the spatial redundancy contained in image data itself. In this work we use image with 24 color bit. VCDCUT used to split the image into the sequence of frames and then split its color (RGB) to (YUV) component. Skew algorithm used to apply motion only on the Y luminance component of sequence of frame, then apply wavelet transform on Components and apply multi multisampling process on (UV) components. The quantization stage has been done to transform image from two dimension to one dimension by run length encoding (RLE) algorithm. In the decompression process the inverse length encoding has been done first and de quantization process has been applied to transform the (RGB) to the (YUV). The inverse of wavelet transform apply on the Y component and de sampling process has been made on the (UV) component then return YUV component to RGB color . Enhancement process done on the result in order to remove noise. Finally PSNR has been computed to check the compression ratio in the resulted compressed image.

The World Wide Web (WWW) has become infinite repository of information over the world because of the vast amount of information that they can gain. Also the access to the web enables any user easily to get information anytime and anywhere. But with these advantages, there is also some disadvantages such as the some websites contains harmful material like terrorism, hate, violence, and pornographic. Therefore developing a system that can filter these contents is become necessary. In this paper we introduce a web content filtering using multi agents, in this system we used FCM algorithm and some webpage feature in classification process. Experimental results based on (2000) different websites are given us an efficient results. On the testing set, the system was able to achieve correct acceptance and reject of (97.5). English pornographic websites uses as a case study.

The Internet has fundamentally changed the way we collect, access, and deliver information. However, this now means that finding the exact information we need is a significant problem. While search engines can find information based on the keywords we provide, using this technique alone is insufficient for rich information retrieval. Consequently, solutions, which lack the understanding of the syntax and semantics of content, find it difficult to accurately access the information we need. New approaches have been proposed that try to overcome this limitation by utilising Semantic Web and Linked Data techniques. Content is serialised using RDF, and queries executed using SPARQL. This approach requires an exact match between the query structure and the RDF content. While this is an improvement to keyword-based search, there is no support for probabilistic reasoning to show how close a query is to the content being searched. In this paper, we address this limitation by converting RDF content into a matrix of features and treat queries as a classification problem. We have successfully developed a working prototype system to demonstrate the applicability of our approach.