MSIT Theses and Dissertations (2018)
Permanent URI for this collection
Browse
Browsing MSIT Theses and Dissertations (2018) by Title
Now showing 1 - 15 of 15
Results Per Page
Sort Options
- ItemArtificial exam scorer for efficient marking and grading of short essay tests(Strathmore University, 2018) Menya, Edmond OdhiamboLearning is an integral aspect to the development of students as well as progressing of a society. The process is always marked with milestones from class work to semester projects and eventually examinations. Students are always required, as a standard, to sit for an instructor set exam paper. The grade and scores that the student garners indicator of progress, amount of knowledge acquired as well as whether or not the student is qualified for the next academic level. Exams are thus an imperative aspect in the academic life cycle and a critical one for that matter. However, the examinations marking and grading process has been marred with inefficiencies, irregularities and unethical practices over the years. This study aimed at achieving the automation of the exam marking process. This approach seeks to introduce efficiencies cutting down time and cost involved in examinations marking in addition to eliminating human bias in the marking process. Research objectives were centered around studying accuracy levels of past exam papers marked by human instructors, reviewing challenges linked to the examination marking process, reviewing existing models, frameworks, architectures and algorithms that have tried exam marking automation, to develop an improved algorithm-based solution that is efficient for the marking problem and performing of experiments to validate the algorithm. The research engaged experimental research experimenting the relation between keywords, synonyms and their related words involvement in artificial marking and marking accuracy. The outcome is an algorithm that mines related words and counts between scheme and student answer to mark exams. The findings were that the model achieves an improved marking accuracy by a margin of 16% from 73% to 89%. The model achieved more accuracy when grading lower mark answers achieving 99.9% when marking 1-mark answers.
- ItemCirculating tumor identification using neural networks for monitoring cancer progression(Strathmore University, 2018) Obonyo, Stephen OketchCancer is the third most killer disease in Kenya after infectious and cardiovascular diseases. It contributes to a significant portion of annual national deaths, led by breast and prostate cancer. Existing cancer treatment methods vary from patient to another based on the type and stage of tumor development. The treatment modalities such as surgery, chemotherapy and radiation have been successful when the disease is detected early and constantly monitored. Ineffective treatment method and development of complications such as cancer relapse must be monitored as they are likely to cause more deaths. Detection of circulating tumor cells (CTC’s) is a pivotal monitoring method which involves identification of cancer related substances known as tumor markers. These are often excreted by primary tumors into patient’s blood. The presence, absence or number of CTC’s can be used to evaluate patient’s disease progression and determine the effectiveness of current treatment option. This research work proposed an adaptive learning-based, computational model to help in cancer monitoring. It identifies and enumerates CTC’s based on the auto-learned features from stained CTC images using deep learning methodology. The 3.0% error rate model, without human intervention, automatically learned the best set of representative features from labelled samples. The representations were used in enumerating and identifying CTC’s given a new test example.
- ItemCloud based prototype for electronic tea auctioning: case of Momul tea factory(Strathmore University, 2018) Chepkwony, Irine CherotichKenya is among the global leaders in tea exports. Tea sub-sector is among the largest foreign exchange earners in the county. Processed tea is sold through the Mombasa tea auction; each tea factory is represented by a broker. The method of auction is “open cry “; all buyers and sellers meet and compete for tea sales and purchases publicly; the prices and awards are also public. The auction processes are currently done manually and the auction model is broker centric. There have been serious challenges affecting the Mombasa auction house; inefficiency in the manual processes, lack of transparency to all stakeholders and concerns of unfair auctions practices of collusions and price manipulation. The aim of this research was to remodel tea auction using information technology innovation as an agent of that change. The remodelling involves devolving tea auctions so that the tea processing factories can auction their own tea directly to the buyers. This is achieved through the use of a cloud-based e-auctioning prototype. By enabling factories auction their own processed tea, it resolves the issues of transparently, collusions and price manipulations and it indirectly reduces the cost of production through minimal dependence on brokers for tea auctions thus positive impact on raising the overall income. The research design used is a mixed method research design; qualitative and quantitative, information was collected from written literature; raw data were collected from the population of the case study: Momul tea factory limited by means of interviews and questionnaires and were analysed using SPSS. The analysis results were presented in form graphs of which over 80% of the respondents felt that brokers controlling the auction process were the main challenge ailing tea auction in Kenya. The tools used to develop the prototype are: MySQL, PHP, HTML and JavaScript, these are open source programming tools. Usability tests were carried out among the selected users. The overall reception of the concept was positive with recommendations to include more factories and tea trade regulatory bodies.
- ItemFall army-worm prediction model on the maize crop in Kenya: an internet of things based approach(Strathmore University, 2018) Ateya, Shantal MusunguIn March 2017, the agricultural sector in Kenya experienced a FAW pest infestation that resulted in the loss of agricultural yields amounting to millions of shillings. The fall armyworm pest caught farmers and agricultural organizations by surprise when it hit most major maize farming regions in Kenya. Currently, both large-scale and small-scale farmers rely on manual observation of the maize crop for detection of the FAW. This comes weeks after the pest has fully matured and began causing damage to crops. The late detection of the FAW in turn results to delays in administering effective pest control measures which forces farmers to incur high costs in administering appropriate control measures. With the ineffectiveness of late manual observations, there is need for an early technology-based solution that will allow farmers to prepare in advance for possible FAW infestations This study proposes the development of a prediction model of a FAW invasion using Internet of Things and machine learning techniques. We suggest the development of a model that automatically predicts a possible invasion by the FAW based on several factors. The key parameters used in the study will be soil temperature and humidity collected through sensors placed in the maize fields. These factors favour the development of the pupa stage of the FAW which later matures into moths that fly to different fields. Based on the parameters, the model will be able to detect the presence of FAW pupa in the soil and issue early warnings to farmers thus allowing for preparation and appropriate counter measures. The study will provide performance evaluation of the model based on the accuracy of the classification, the precision and recall ratio of the collected parameters. The developed model achieved an accuracy of 82.06%.
- ItemFuzzy expert based real time monitoring system for patients with chronic heart failure through IOT(Strathmore University, 2018) Muriuki, Isaack MwendaData from the World Health Organization has placed CHF as the number one global killer. It remains the only cardiovascular disease with an increasing hospitalization burden and a continuous drain on health care budgets. Heart failure is a complex clinical syndrome of symptoms that suggest the heart is unable to pump blood efficiently as it should. Heart failure signs and symptoms may include irregular heart rate, blood pressure, fatigue and weakness. The hard reality, with which doctors contend every day, is that the effects of these conditions often manifest too gradually for people to recognize. It falls to the healthcare system to deal with these diseases after they’ve advanced to a serious stage, often at a great financial cost. Effective therapy and treatment in CHF patients require thorough continuous monitoring of patients vitals. Doctors require information on patients; blood pressure, heart electrocardiography activities heart rate and temperature to predict the heart failure attacks and respond swiftly. The typical way to diagnose and monitor CHF patients is by use of bedside patient monitoring systems which requires monitoring within the confines of the hospital. Such monitoring equipment are available in very few hospitals in Kenya and that is an impediment to proper therapy and treatment for CHF patients. The challenges faced in using the existing methods include; lack of flexibility for the patient as there is need for long term monitoring in a hospital setup, financial burden on the patients when they are hospitalized, obtrusive nature of the current monitoring systems making it not suitable for monitoring outdoors. This research applies scrum methodology to design, develop and test a fuzzy based expert system for real time monitoring of chronic heart failure patients through IoT. The IoT architecture contains sensors to capture heart rate, heartbeat, and temperature values from the patients and transmit values from the Arduino board to an IoT server via a GSM communication module. A mobile application will be developed to enable the care givers to monitor the patient remotely. The recommended vital parameters will be keyed to the system to enable it detect anomalies. As a result, patient's doctor and care-givers can see CHF patients vital current health conditions in real-time and get sms alerts in case of anomalies’ to enable them respond swiftly. This model reduce re-hospitalization, enables adjustment of therapy to accommodate change in the patient’s condition and reduces death rates caused by CHF.
- ItemMhealth application for community health volunteers data collection: a case of Makueni county(Strathmore University, 2018) Matheka, Morris MuthusiThe Community Health Strategy in Kenya was introduced to strengthen linkages between communities and the formal health system. However, one of the major drawbacks has been reporting. The manual process for capturing data using Ministry of Health registers by Community Health Volunteers have presented challenges in accuracy, completeness and timeliness of health data from Community Units which has a negative impact in health service delivery due to the delays in quality data reaching the health facilities. The healthcare industry is undergoing a major paradigm shift due to the rapid advances and developments in mobile technology, mHealth that use mobile devices and other wireless technology in medical care. This research focuses on using the dynamic system development methodology to develop a mobile application that will be used by the Community Health Volunteers to collect health data from the household. Data collection will involve both primary and secondary sources which will include literature review and questionnaires. The findings of the research established that the application addresses challenges faced by the manual data collection, the manual register takes 4 weeks be delivered to the health facility using the mobile application it takes only minutes for the data to be availed. Above 93% the respondents indicated that the application assists in improvement in accuracy and 100% of the respondents indicated that the application contributes in collecting complete data, thus contribute to improvement of service delivery to the community. The application should be scaled-up and deployed to all counties in the country.
- ItemMobile application for filing of and payment for Intellectual Property Rights using QR code: case of Kenya industrial property institute(Strathmore University, 2018) Andati, Eric MalobaEnsuring secure transmission of sensitive data and payment of transaction fees has been one of the challenges affecting customers and businesses. Intellectual Property (IP) field is one such area that has faced such challenge. Over the years, IP has grown in importance, attracting greater interest and increased need by inventors and other IP rights holders to seek protection of their inventions and other IP rights. To ensure protection of these rights, applicants are required to file their applications at IP offices and remit various fees during the examination process, as well as pay annual maintenance fee for the protection to remain valid. While filing for IP rights, applicants face security challenge, as their IP data can be intercepted while in transit or be exposed to third parties thus compromising their inventions. In addition, while making payment of IP fees, they face challenges such as delayed transactions and platform incompatibility. On the other hand, IP offices are susceptible to loss of revenue as a result of less-than-secure payment methods used. Hence, this study aimed at establishing how proximity/contactless technology could be incorporated into mobile-based devices to support secure mobile filing of and payment systems for IP rights. This research therefore proposed a process to develop a QR code-based mobile application that would facilitate speedy and secure filing and transmission of IP data as well as settlement of payments by IP rights holders to IP offices. Consequently, a functional mobile application that can generate a QR code, post the same to a remote server and make payment by scanning a QR code is presented. Additionally, a simple web page is provided to present the submitted information which has been encoded in QR format. Data collection was achieved by means of questionnaires and review of secondary data sources. The study was conducted in line with ethical practices as specified by the University rules and regulations.
- ItemA Mobile application for smart-shopping for retail stores (M-kiosk)(Strathmore University, 2018) Thuku, James GichuhiMajority of retail stores in Kenya commonly referred to as kiosks are yet to automate their operations. The mode of operation for majority of the kiosks is manual. This brings about several challenges to both the store owner as well as the customers. Some of the challenges faced by the retailer include lack of instant information regarding products and customers, human error during audit of transactions and tracking of stock is very slow. For the customer, manual operations are prone to long queues, slow to response on product related queries, and no direct interface to interact with the retail store at their comfort. With increase in competition, there is need to get innovative on the use of customer information to enhance customer satisfaction. There also exists no mechanism to reward customers who are loyal to the shop. Availability, performance, productivity, cost reduction and reliability are good reasons for considering an automated solution. This research proposes a model for automating operations and analysis of customer data in a retail store setting to enable the business to run efficiently. A mobile and web based prototype was designed, developed and tested to assist in automating operations, analyzing customer purchases to increase efficiency. Products are assigned a unique bar code which is scanned during purchase, allowing the shopkeeper to easily track stock, and group customers into different categories based on similarity of purchases. These categories will be used to customize promotions to specific customers. The study will identify relationships in customer purchasing habits with an objective of deriving usable values through increasing sales and enhancing customer loyalty and satisfaction. The store owner will use his mobile phone to scan the product as well as the customer card where applicable, and tally the total amount to be paid for the products. The customers are able to interact with the system, and view past transactions as well as get notifications regarding promotions using a mobile based application on their phone. The design of promotions will be done from the backend, and will be based on frequently purchased goods, hence achieving customized promotions. The study has adopted concepts from Social Network Analysis (SNA) , and applied the same to enable retail stores experience Business Intelligence (BI) through studying human behaviour with an aim of creating more opportunities for them to reap more benefits by using the data previously thought to be of no help.
- ItemMPLS (Multi-Protocol Label Switching) assisted routing procedure in Software Defined Networking (SDN)(Strathmore University, 2018) Otieno, Humphrey OwuorMulti-protocol label switching has been incorporated into provider networks to provide quality of service. Owing to the design of the protocol, its ability to push and pop labels in packets, independent of their underlying protocol makes it popular in interconnecting multiple networks in to one transport pipeline. At the same time, multi-protocol label switching has proven to be a very fast procedure for forwarding devices because the central processing unit cycles required in making a forwarding decision is far less compared to traditional forwarding decision-making metrics like analyzing the internet protocol header. However, current multi-protocol label switching implementation is a complex configuration procedure and does not provide a central bird’s eye view of the network topology to network engineers. Logging in to every label switching router and loading multi-protocol label switching configurations to allow it to connect to neighboring label switching routers in the label switching path is required. Allowing network engineers to have a central view and control of the network topology while still providing multi-protocol label switching services in a simplistic approach will make them achieve adaptive routing and traffic engineering seamlessly. This will improve quality of service and quality of experience in transport networks. Software defined networking is the approach this research takes towards providing central control because of the flexibility, programmability, and adaptability of the technology. This work proposed the design of a routing procedure that will implement multi-protocol label switching on a software defined network via OpenFlow. Experimental synthesis and prototyping approach was used to achieve the research objectives. A simulated environment called Mininet provided the implementation test bed. Internet control message packets were the test data to show how multi-protocol label switching labels are added and stripped. An illustration of the packet capture information from the experiment was presented and analyzed.
- ItemA Plasma glucose prediction tool based on dietary assessment: a case of type 2 diabetes patient(Strathmore University, 2018) Njihia, Alex WainainaManagement and control of blood sugar using dietary intervention has for a long time been considered to be important. The caregivers have always advised diabetic patients to moderate the amount of carbohydrates intake. The approach here has always been reduction in the amount of carbohydrates, unfortunately this does not translate to the reduction on the blood sugar in some case. This is explained by the fact that what determines the sugar levels in the blood has to do more with the glycemic load of the carbohydrates consumed which is dependent on the glycemic index of the food item consumed. Though the amount of carbohydrates taken by the patient has a role to play, it is rather indirect. The study, sought to develop a tool for the computation of the glycemic load of the food item consumed by an individual by aggregating the various meals parameters. The tool has been developed by analyzing the dietary factors that affect the glycemic load and using these factors has the regressing variables. The algorithms used in the development of the dietary assessment tool have been used to map and mine the standard glycemic index of individual food item and to estimate individual patient meal item glycemic using regression analysis approach. Experimental data results indicate the tool can compute the glycemic load of the food item which is comparable to the standard glycemic load values and it also gives plasma glucose prediction trajectories which mirrors those obtained from existing clinical trial dataset.
- ItemA Prototype for mapping of tweets on state services for decision support: a case of Huduma Kenya(Strathmore University, 2018) Ng’ang’ira, Judy NyakairuThe growing public participation in decision making regarding the management of State resources demands for a tool that support meaningful insight of the many aspects on environmental issues, for the development and evaluation of alternative management options. Twitter has become quite popular among researchers due to its massive volume in data thus drawing a great interest by the public service community to answer questions relating to the use and misuse of public service offices. However, despite the growing participation by the researchers using twitter as a public service misuse detection the State does not seem to optimize the opportunity that twitter offers for detecting and monitoring the services offered at Huduma Kenya, a one stop shop offering a variety of state services in almost all the counties. The main objective of this research was to demonstrate that twitter tweets can be dependably grouped based on state services selected keywords. The magnitude of state service tweets can be predicted with high accuracy. The method used includes various steps that can be summarized as first categorizing the groups on twitter and defining them. Second, finding out how each group pattern of activity contributes value in group participation. Thirdly, the identified users were invited to contribute in the interviews. Fourth, analysis of the interview results was carried out enabling the researcher to identify findings of ill-structured decisions in state services. Also, a mixture of related investigation similarity diagramming and grounded theory techniques were used to identify different benefit-related trends, patterns, and evolving relationships through all interviewees. Then, the data was sorted and compared by group type to discover which themes were most repetitively related per group. Moreover, to estimate on the generalizability of these results to the user population at large, access usage logs was required to determine usage levels.
- ItemA Prototype to forecast trends in rainfall amounts in Kenya(Strathmore University, 2018) Onyango, Mary MargaretFood security has been a key subject of interest to governments, non-governmental organizations such as the World Food Programme (WFP), and the Food and Agriculture Organization of the United Nations (FAO), and the local population worldwide. One of the growing problems in the food security arena is the continual rise in the number of people who are food insecure in recent years. Previous solutions have included government interventions through food pricing and agricultural subsidy policies, and non-governmental interventions through harmonization and distribution of information on food security. However, the need to come up with more and better solutions to the growing problem has been emphasized, with calls to different groups and individuals being invited in recognizing that food security is a shared responsibility. The proposed research is an application research, which employs the use of a prototype to show rainfall trends in rainfall. In a bid to reduce the time taken to access data, the research seeks to employ the use of existing data which can be accessed from online sources, with deliberate sampling employed as the main data collection method. Data sets will be sourced from the World Bank online portal. The research will then recommend the proposed prototype as a base for the measurement of rainfall that will influence decision making on food security at a local, national and regional level to local and regional food security institutions, the government of Kenya and other interested organizations.
- ItemA Real-time location based algorithm for notification of crime hot-spots using crowd sourcing(Strathmore University, 2018) Chepngetich, MarylineSecurity of the people has always been the number one objective of many governments in the world today. Governments endeavour to achieve this objective has faced several challenges ranging from economic, social and political. Despite heavy investments by local and National Government in Kenya on security measures, crime continues to remain a serious problem in the society, as a result, there are loss of lives, loss of property and investors shying away. Gathering relevant and up to date operational information on crime intelligence across several sources has always been one of the challenging issues faced by national security practitioners and citizens. This therefore makes it difficult to identify crime hotspot areas in timely manner, and also improper allocation of Police resources in the right hotspot areas. The data collection exercise was done earnestly to ensure that there was ample understanding of the participants’ interaction with crowdsourcing platforms and their experience and willingness to use a crowd-based crime hotspot reporting network. The study thus found significant justification for the design of the criminal hotspot system to leverage data about crime incidents in the city in order to classify crime hotspots. The design of the system was made using unified modelling language and detailed in the fifth chapter of the thesis. The developed prototype was then tested against parameters to gauge its efficiency and effectiveness. The conclusions of the testing as well as the recommendations of the study are documented in the sixth and last chapter of the study respectively.
- ItemA Secure electronic document management system using public key encryption: a case of Strathmore University(Strathmore University, 2018) Momanyi, Stephen BichageThe need to safeguard and ensure the authenticity of academic records is paramount for any reputable educational institution. As institutions are moving into the digital era, more and more documents are being stored in digital form. This comes with its own challenges and is compounded by the fact that computer files can be modified without leaving any trace, and one cannot tell the difference between the copy and the original. Hence, the need to employ security measures, such as cryptographic techniques to ensure the integrity of digital documents. Paper documents have been used as backups of academic information store in academic management systems, these are difficult to manage and entail a great deal of manpower to maintain. Besides, they also require large storage facilities. The volumes of the paper documents have increased over the years and this has necessitated the search for a better way of managing them. A lot of effort has gone into research of verification of academic certificates, while little has been done ensuring that accurate grade information maintained in academic management systems. Strict procedures must be put in place to ensure there is not tampering of the grade information in the academic management systems. In this regard, a secure document management system using public key encryption can provide a solution for storage and security of digital documents. In this research, we explored how these systems can be used to protect the integrity of digital documents. This research proposes the use of a secure document management system using public key encryption that will facilitate the digital signing and storage of the digital documents. To implement the model, we developed a system that manages user rights and enables authorized users to use a public certificate to embed a digital signature on digital documents. The system was implemented using a document management system with an additional module for digital signing. Private keys and digital certificates were generated for each of the authorized user and uploaded to the system using each user’s credentials. To apply their digital signatures, the users’ must input a password to complete the process, thereby providing an addition layer of security. The system was tested for functionality, usability and effectiveness in managing digital documents and from the test results it shows that the system can be relied upon to securely and efficiently manage digital documents.
- ItemStock market price prediction using sentiment analysis: a case study of Nairobi stock exchange market(Strathmore University, 2018) Lwanga, Victor KwomeStock market price prediction has become an area of research and interest for several years now due to the many challenges in making accurate price predictions due to the volatility of the data. However, the stock market is not easily predicted. Movement in the stock market is influenced by various factors such as personal fortunes, political events, individual tastes, preferences and natural disasters. People can express all these through their sentiments and opinions on the social media platforms, financial news, and blogs. The stock price does not only rely on the law of demand and supply. People’s opinions and moods also have a substantial impact on the movement of the stock prices of a company. Recently, efforts to increase the accuracy of stock market predictions by including data from social media such as Facebook and Twitter has received a lot of attention. Social media can be regarded as an indicator of sentiments, and these are known to influence the stock market. Current models lack a clear interpretation, and it is also difficult to determine, which data is relevant for stock market prediction since there is an abundance of the same on social media. This study proposed the use of machine learning algorithms that will be utilized in Natural Language Processing (NLP) to get opinions and sentiments from social media on a particular company's stock to predict the stock market prices. Previous studies show that public mood, opinion, and stock market price have some relation to an extent. The research used Support Vector Machine with bigram feature to perform sentiment analysis which exhibited and accuracy of 83 percent and Artificial Neural Network in Stock price prediction which had a mean squared error of 5.6. This research has proven that sentiment analysis can be incorporated in stock price prediction.