Browse
Recent Submissions
Item A content base technique for the detection of sms spam data(UMT.Lahore, 2018) Muhammad FarooqToday text messages have a significant impact on our lives, but at the same time we face many critical problems due to SMS spamming. Therefore to detect spam messages and distinguish it via accurate filtering is a challenging task for researchers. In current research work content based spam SMS filtering technique is proposed depending on machine learning approach to distinguish spam messages firom mobile data while considering the low processing power and limited memory of mobile phones. From each text message; five attributes are extracted and based on these attributes/features, an unknown/unlabelled SMS message can be classified as spam or ham by a trained leaming algorithm. These attributes are the length of the text message and presence of repeatedly occurring spam words, count of spam words, combination of spam words and SMS class. It is shown that Decision Tree classifier performance is better than other machine leaming algorithms investigated. The other learning algorithms explored in this work are Naive Bayes and neural network architecture-Multilayer Perceptron AlgorithmItem Prediction of carboxylation lysine sites using position relative features and statistical moments(UMT.Lahore, 2018) SABA AMANATCarboxylation is the type of post translational modification that takes place at lysine. Lysine residues play an important part in catalytic reactions, calcium absorption and construct muscle protein. For studying the biological function, it is important to correctly determine lysine sites sensitive to carboxylation, but it is costly and time consuming process to determine them experimentally. On the contrary, the reliable computational model is required for identification of carboxylation lysine sites. In this paper, we present a computational model for the prediction of the carboxylation lysine site which is based on machine learning. Training of the model is done by neural network using experimentally verified and updated data. Statistical moments have been used to train a neural network. The model is validated by jackknife, cross-validation, self-consistency and independent testing. A test revealed 93.93% for jackknife test, 95.16% for cross-validation, 94.3% for self-consistency and 94.3% for independent testing. Proposed model has 78.26% sensitivity, 96.4% specificity, 95.16% accuracy and 0.6776 Mathewās correlation coefficient, it reveals that the proposed model has better performance as compare to existing model PreLysCar.Item Identification of lysine methylation sites using position relative features and statistical moments(UMT.Lahore, 2018) SARAH ILYASMethylation is one of the most important Post-Translational Modification in human body which usually arises on lysine among the most intensely modified residues. It performs a dynamic character in numerous biological procedures such as regulation of gene expression, regulation of protein function, and RNA processing. Therefore, to identify lysine methylation sites is an important challenge as some experimental procedures are time consuming. In this research, a computational method is used to identify lysine methylation sites. Firstly we construct feature vectors by using some feature extraction techniques including position and scale- invariant features. The key discriminating attributes in the feature vector are site vicinity vector, raw, central and Hahn moments of position relative incidence matrix and reverse position relative incidence matrix along with the two dimensional primary structure. A neural network is trained based on the extracted features. The performance of the proposed method is then validated using cross validation and jacknife testing. The evaluation of the performance is calculated by using performance metrics including Accuracy, Sensitivity, Specificity, Mathewās correlation coefficient. The results obtained from these metrics are 96.7%, 95.8%, 97.7%, and 0.9345 respectively, which shows that the model outperforms to identify lysine methylation sites than the existing ones such as iMethyl-PseACC, BPB-PPMS and PMeS.Item Under-graduate study recommendations using association rule mining(UMT.Lahore, 2018) Hira AsimIn this research work, we are aimed to provide recommendations regarding degree programs and institutes at under-graduate level to students who have successfully passed their intermediate. In order to recommend we collected data via Google survey form regarding background study area, future goals and interests of individuals who are successfully done with their under-graduate studies for least. After the data has been collected it was being pre-proceed, visualized and analyzed for future use. Data mining technique named Association Rule Mining is applied on the data to find hidden patterns and trends in it. The results of the Association Rule Mining technique are the recommendations to the students regarding suitable under-graduate degree programs and educational institutes as per their background study stream, area of interest and future goals. Recommendations are based on the information collected where students and graduates have alike choices regarding study streams, area of interest and future goals. We are hopeful that this contribution will be a huge step forward towards the betterment of educational environment as well as a source of proper guidance for the students who are done with their intermediate and now aimed to take admission in an under-graduate program in a well reputed university in Pakistan.Item Activity recognition in smart homes(UMT.Lahore, 2018) HINA JILLANIRecognition the activity of daily living and in smart home environments is one of the challenging tasks because it depends upon the human behavior and on different sensors. In this thesis, the algorithms that are being presented are neural network multi-layer Perceptron (MLP), Support Vector Machine (SVM) and NaĆÆve Bayesian for automatically detecting the activity in smart homes. The SVM implemented on that dataset by using different kernel and simple method of SVM and shows the results that which kernels is best and gave accurate results. These algorithms are implemented on the dataset of MIT lab that have captured the activity of daily living for recognizing the activity in the smart home environment. The result of these techniques (MLP, Naive Bayesian, and SVM) is being compared with each other and results shows that the NaĆÆve Bayesian is best suited for that data.Item Blockchain technology for healthcare systems(UMT.Lahore, 2018) MUHAMMAD USMAN FAIZHealth information management systems are responsible for managing health records digitally in a manner of sharing and storing records. As health informatics is a very immense domain with containing very complex data in very huge amounts. This diverse nature of health data makes it very much difficult to perform efficient sharing and storing. In this research I have proposed a way of storing such health related data using blockchain technology. With different variations in blockchain frameworks this research finds out the suitable blockchain for storing and retrieving health data. This research gives a complete storage model including blockchain and distributed database for faster and efficient retrieval of data.Item Data center tier design and efficient Techniques of core routing(UMT.Lahore, 2018) ATIF MANZOORData Center is a repository which is composed of IT equipment which includes Servers, Storages, and networking equipment, i.e. Firewalls, Core Routers, Switches and Racks to be integrated IT equipment. In the modern era, data center plays a vital role in IT based organization to process, store and distribute huge volumes of data. In the recent technology, data center is designed for centralized architecture of high availability to maintain IT services up and running for an upright business relationship to the customer. Routing is the process of data path selection of IP networks. Routers perform path selection of the basis of routing tables stored in their memory. Routing- table contains IP routes for route transformation via the best path in the networks. Service providers use different routing protocols in their enterprise networks. These routing-protocols have the limitation of non-convergence in the networks. Route redistribution is the technique which overcomes this limitation. Due to this technique, service providers can get optimized communication with IP networks where multiple routing protocols is being used. This research thesis describes: (i) How to design state-of-art Data Center and (ii) Efficient techniques of core routing using different metrics i.e. Convergence, Throughput and Packet Delay for better performance of enterprise networks. So, with the help of this research thesis, data center cost has been minimized, proper tier standards have been followed and performance efficiency in enterprise networking has been improved using core routing.Item Fpgas and gpus as performance accelerators in cloud and large scale datacenters(UMT.Lahore, 2018) HAMID HASSANField Programmable Gate Arrays (FPGA) and Graphical Processing Unit (GPU) are being widely used as high performance processing units due to their speedy processing, parallelism and programmable flexibility for customized tasks. Due to the high computational requirements of Cloud based infrastructure, FPGA and GPU are now being used in Clouds as performance accelerators. This Research work presents comparative analysis of FPGA & GPU based solution by simulating in large scale infrastructure. This study will help to adopt novel techniques for performance acceleration in cloud and large scale infrastructure, and proves to be cost effective solution.Item A survey of resource allocation techniques in cloud computing(UMT.Lahore, 2018) MUMAMMAD FARAZ MANZOORCloud computing has become a very important computing model to process data and execute computationally concentrated applications in pay-per-use method. Resource allocation is a process in which the resources are allocated to customers by cloud providers. As the data is expanding every day, allocating resource efficiently according to the consumers demand has also become very important, keeping SLA between service providers and consumers in prospect. Due to the finite resources available, it has become difficult for the service providers to allocate the resources according to the demands. Many unique models and technique have been proposed to allocate resources efficiently. In light of the uniqueness of the models and techniques, the main aim of the resource allocation is to limit the overhead/expenses associated with it. The absence of a comprehensive survey covering the aspects of strategic, target resources, auction, optimization, scheduling and power has encouraged a survey of existing cloud resource allocation techniques, this paper reviewed and analyzed the current state-of-the-art cloud resource allocation techniques. Also, a topical taxonomy is exhibited in light of resource allocation enhancement objectives to arrange the existing workItem Increasing students involvement by gamification in assessment evaluation(UMT.Lahore, 2018) Qasim AliIn universities, studentās registers in many courses and they have to submit assignments, quizzes and work on subject projects. They find all these tasks old fashioned and donāt entirely corporate and adequately engage with the course materials. Therefore, students donāt get good grades. Secondly, students canāt get feedback on their work on time, and they become confused either they are on the right path or not and they donāt get the chance to check their progress at any time. This study aimed to gamify course content in such a way that student motivates and participate in classroom activities and manage their time and give them quick feedback and progress report. This research generates a gamification toolbox by blending course outline, project-centric approach, and gamification techniques by developing an online web portal and badges, leaderboard, SMS and web notifications are its main features. This study took two model classes into its observations and performed a series of test on the data. Student t-test shows the level performed better where gamification toolbox was applied. Correlation between project deliverable and questions in exams related to same domain show a strong relationship in the class where gamification of this research was involved.Item Predvir(UMT.Lahore, 2018) ANUM SHAHZADIProteins are typically virulence factors of bacteria and other molecules that are produced by enzyme. Virulent is a bacterial microorganism that a reason to cause the disease which is describe the term of infecting bacteria that is involve through the host body and natural bacterial virulence elements. Prediction of virulent protein datasets is used for the recognition and classification of the virulent related features. That is used to discover the drug, vaccine against proteins necessary for pathogenicity. It is highly desired to build a computational model that can identify the virulent and non-virulent protein sequence accurately and efficiently. This study proposed an efficient method that is based on the classifier for the prediction of virulent proteins. Using Statistical moments, the features were extracted from the benchmark dataset that comprises 1193 virulent protein and 2179 non-virulent proteins. The predictor was implemented using neural network model. Highest accuracy provided by the neural network was 89.3%, using independent dataset testing. The results produced by the trained model on independent testing were validated using 10 fold cross validation and jackknife testing. 10 fold cross validation results are where sensitivity is 77.89 specificity is 93.62 accuracy is 87.08 and MCC is 7484. All of these are producing better results than the old techniques like SVM. This method is ultimately more accurate, cost effective and high throughput technique for the identification of virulent protein and non-virulent protein sequence.Item Prediction of nitrotyrosine sites based on composition and position based features(UMT.Lahore, 2018) Ahmad WaseemClosely related to causes of various diseases such as rheumatoid arthritis, septic shock, and coeliac disease; tyrosine nitration is considered as one of the most important posttranslational modification in proteins. Inside a cell, such modifications occur accurately by the action of sophisticated cellular machinery. This task is accomplished by specific enzymes present in endoplasmic reticulum. The identification of potential tyrosine residues in a protein primary sequence which can be nitrated is a challenging task. To counter the prevailing, laborious and time-consuming experimental approaches, here we introduce a novel computational model. Based on experimentally verified tyrosine nitration sites, they are transformed to their feature vectors. An adaptive training algorithm is then used to train a back propagation neural network for prediction purposes. To objectively measure the accuracy of the proposed model, rigorous verification and validation tests are carried out which led to a promising accuracy of 88%, a sensitivity of 85% and a specificity of 89.18% and Mathew correlation coefficient of 0.627. We believe that this computational model may provide foundation for further investigation and can be used deal with the other PTM sites in proteins.Item Sproteasepred(UMT.Lahore, 2018) NAJM AMINAmong all the protease prediction of protease from proteins is the most important and use for highly regulatory physiological processes. Protease performs different functions on the base of different biological processes also classify this protease in five different types. Identify through the sequences that its protease or non-protease. Itās necessary to recognize the protease urgently in different parasites, viruses, and bacteriaās also in the growth, birth, digestion, and maturation and death organisms. Neural network is a predicted method which provider fast and accurate results as compare to the old models like MIEC-PLS. Neural network is introduced to address this problem and avoid the redundancy. Overall success full rates are obtained by cross validation and independent testing for protease and non-protease is 91.8% and identifying the protease types are 82.0% among the five typeās aspartic, cysteine, serine, metallo and threonine. These high results indicate that this predicted model is power full and become a convenient tool in bioinformatics.Item Sentiment analysis of user generated text (movie reviews) using machine learning(UMT.Lahore, 2018) Umar RashidThe film industry play a vital role in the business, entertainment and employment. The rapid and exponential growth of text on web has introduced the new horizon of analysis on the sentiment of user generated text. The users are most interested to give feedback or comment using emails, reviews, social media networking sites, etc. Our research focus on the users based view in which we present the new method to extract and analyze the adjective. The adjective is main part for exploiting the results of sentiments. Opinion analysis helps to identify the polarity of the text and feature extraction. This research is an attempt to provide an efficient framework to calculate the sentiments of written text by using the machine learning techniques Support Vector Machine and Random Forest. For sentiment analysis large data set of 50000 movie review is used. The results achieved good accuracy of 95.3 percent.Item Prep4Job(UMT.Lahore, 2018) Makhdoom Muhammad Imran; Shabaz Khalid; Farhan MaqsoodPrep4Job is platform which provides a complete set of features for recruitment related processes. User will be able to access all the features in a single platform as no website that is providing an entire set of features for recruitment related processes. The system includes two main categories of stakeholders which are candidates for jobs and employers, both having different specs, features and constrains to use the platform. Features like Dashboard Management, Online Video Conversation, Cloud Data Storage, Interview Management and Gateway Payment are the main features of this platform. A user will be able to set up a company profile or a job seeker profile where they can view the list of users seeking for jobs or list of posted jobs related to the company or university in which they are interested. User will be able to search, view, apply and post jobs, provide feedback in order to rate other user, dynamically manage profile along with their dashboard and interview management dashboard. The administrator will have complete privileges.Item Context aware news similarity for the identification of follow-ups among human loss related news articles(UMT.Lahore, 2018) Waqas AliText similarity plays an important role in document clustering, plagiarism detection, automatic student answer grading, information retrieval and language translation systems. Many researchers have studied on string, corpus and knowledge based approaches to resolve the problem of document similarity. In this paper research has been made on full text similarity and context information of a news. Time complexity and follow-ups distribution are main challenges for full text similarity which is computed by using jaccard index. The proposed system in this paper uses context information of news to resolve these issue and to classify the news into three categories i.e. same day follow-ups, different day follow-ups and distinct news. Context is built by using well known Stanford parser which further enriched with factor of severity to reduce the target dataset for similarity computation. Results showed that context aware similarity approach is better than traditional full text similarity approach in efficiency and accuracy.Item Text detection and recognition from images or videos(UMT.Lahore, 2018) Faraz AhmedThis paper analyzes and compares problems, challenges, proposed methods, limitations and performance of text detection and recognition from images or videos in last eight years. It summarizes different techniques to detect and recognize text including structure based partition and grouping, key text points , non-text filtering and connected components, energy minimization based framework, gradient and geometric based algorithm, learning based methods and optical character recognition based software. It also addresses state of the art challenges including horizontal and non-horizontal text, text having complex backgrounds, text having different sizes, language independent text, low resolution images having text, curved text lines and handwritten text. This survey shows performance comparison on different benchmark datasets.