MSIT Theses and Dissertations (2020)
Permanent URI for this collection
Browse
Browsing MSIT Theses and Dissertations (2020) by Issue Date
Now showing 1 - 20 of 23
Results Per Page
Sort Options
- ItemA nutrition information dissemination platform for primary caregivers in Kajiado County(Strathmore University, 2020) Wanjau, Samuel MwangiAccess to information by Primary Caregivers for execution of their caregiving role still remains a challenge. These challenges such as lack of contextualized and personalized Nutrition Information is attributed to weak structural and operational initiatives that have been put in by the Government of Kenya and their implementing partners. The overall aim of this study was to develop a nutritional information dissemination platform for primary care givers using USSD technology to improve nutritional education. This platform assisted in tackling the nutritional information gaps that Primary Caregivers have. The researcher employed Agile methodology and Object Oriented Analysis and Design (OOAD) that realized the USSD enabled Nutrition Information Dissemination Platform. The platform was developed on a web platform that makes use of HTML 5, PHP and MySQL to provide user friendly utilization dashboard for health workers and health stakeholders. The proposed solution availed nutrition information to Primary Caregivers which in turn lead to an improvement in the nutrition status of the Kajiado County residents. The study involved stakeholders such as County Government of Kajiado and a nutrition expert from Strathmore Medical Center who contributed by providing guidance throughout the development the USSD solution for Primary Caregivers.
- ItemDynamic knowledge based authentication model for enhancing security of USSD banking transactions(Strathmore University, 2020) Njuguna, Michael WanumaA large part of mobile banking transactions in Africa are facilitated by USSD technology. In authenticating customers, banks rely on a single security vector: a shared secret such as a six-digit PIN. This mechanism presents vulnerabilities that are commonly exploited to perpetuate fraud. In particular, this study focuses on insider threats, privacy leakage and social engineering attacks. To address these challenges, the study proposes a dynamic authentication model that poses diverse challenge questions based on available customer and transactional data. These challenge questions are unique to a given customer and variable over time making it difficult for anyone other than the legitimate user to deduce the correct response. A test-driven approach was used to guide development with the test scenario increasing in complexity after each iteration. Validation tests show the proposed scheme demonstrably provided enhanced security. The true acceptance score for legitimate users stood at 92.8 percent. As for guessing attacks by adversarial users, the probability of a correct guess was reduced to less than 0.08 percent. Performance-wise, the computational overhead increased by only 22 percent as compared to the classical method. This was sufficiently small as not to be noticeable by a user in real-world deployment. The study points to the feasibility of the model but recommends further research on challenge question generation for even greater security.
- ItemAn agency core banking application using blockchain technology(Strathmore University, 2020) Kamau, John KibathiCut throat competition between various institutions in the financial sector in Kenya particularly banks have pushed them into adopting innovative ways to serve their customers and generate revenue. One of the innovations that banks have come up with is the use of agency banking channels to serve their customers. This channel has proved to be very successful of over the years since its launch and has seen many banks take up the channel. However, this channel has opened up a new avenue for frauds and theft by criminals. This has pushed the banks into adopting new security techniques to combat these criminals. Software related security mechanisms have been installed to curb fraud for example the use of Encryption techniques such as Data Encryption Standard (DES) encryption. Although it has shown great promise in securing end to end connection with the bank servers, Data Encryption techniques have multiple weaknesses that we hope the adoption of block chain technology will resolve. The purpose of this project was to come up with an agency banking application that will use block chain technology to assist banks in curbing the vice of fraudulent transactions initiated by fraudulent cards. The model works by replacing the current architecture is use and replacing it with one that uses block chain technology. In order to secure agency banking transactions, the transactions are stored in an immutable ledger which are then chained together to form the block chain. The block chain forms the database because it stores information in a digital ledger in data structures referred to as blocks. In the model, every bank maintains a copy of this Ledger to form a decentralized system that is more secure as opposed to the existing one that is central and is prone to security breaches. This blocks use hashing algorithm to identify each transaction securely. The network is a permissioned network hence only authorized nodes are allowed to process transactions. The model was tested and a transaction was sent to from an agent’s point of sale network to the peer to peer network to the validating nodes where it was successfully validated and the result want sent to the customer via the agent Point of Sale terminal.
- ItemA machine learning algorithm for predicting wild fire occurrence(Strathmore University, 2020) Otieno, Jack OdungaA wild fire is an unplanned fire that burns in a natural area such as forest and grassland. Indeed, wild fires are destroying parts of the world’s forest coverage, affecting land and killing wild life. Majority of wild fires are caused by human activities and weather conditions. Despite various forest management authorities using numerous methods to detect and suppress fire, there are several wild fires reported around the globe annually. In Kenya for instance, Kenya Forest Services ensures fire detection and detention through the use of ground patrols and fixed stations (fire towers). They also use radio systems, vehicles, motorcycles and even bicycles. These techniques are not working effectively due to inadequate staff and resources to cover the available forest coverage. The research proposed the development of supervised machine learning model to predict wild fires using existing data that was collected from credible climatological sources that included both meteorological sources as well as wildfire databases focusing on content dated from 2000 to 2020. The study utilized data sets from multiple sources including National Fire Danger Rating System (NFDRS), Canada National Fire Database (CNFDB), University of California machine learning repository, and scientifically verified Internet sources. The methodology involved collection of relevant data sets, cleaning and preparing the data, training the models, model testing and validation. The climatological factors were used, as input values and Artificial Neural Network (ANN) implemented to establish prediction model. The model was developed through rapid application development (RAD) methodology. Upon completion it was deployed on a web environment to be used by various stakeholders in monitoring and predicting wild fires by giving a binary output of a yes or no on the likelihood of wildfire occurring. Artificial Neural Network Model was trained and validated using 80% and 20% of the set features respectively. The model gave performance accuracy of 82.69 per cent.
- ItemAn intelligent image processing model for context-aware digital signage: case of apparel advertisement(Strathmore University, 2020) Omogo, Thomas OmondiDigital Signage is a way of presenting content such as advertisements, news, menus, and directions on electronic displays on places with high human traffic, Stadiums, Transport hubs, Malls, Retail Stores, and Notice boards. People have a variety of preferences when it comes to style and design of apparel. In developing advertising content, people feel more engaged when viewing tailor-made content fashioned toward a set of characteristics such are skin tone, age, and gender. The ability of digital signs to detect its contextual surrounding and intelligently display an apparel advertisement has been on-demand to create an autonomous content generator for digital signage. Currently, digital advertising content is developed by a designer who has experience both in computer skills, digital design with creativity and innovation. Content management software preloaded onto the digital screens makes it simple to load and display content. However, small scale retailers who invest in digital signage fail to achieve its full potential due to limited knowledge in creating a context-aware advertisement. The proposed research applies machine learning technique to build a model which captures a potential customer’s apparel features using image processing then display a recommended outfit based on the input features. The result is a digital signage that autonomously creates a context-aware apparel advertisement based on what a potential customer is wearing. The model was evaluated based on accuracy, precision, and recall. Anthropometric measurements coupled with apparel segmentation fed into R-CNN gave the best apparel classification. The model’s yielded an accuracy of 94.39%, with a precision of 0.72 and 0.84 recall, respectively.
- ItemA model for forex market price prediction: case of Central Bank of Kenya(Strathmore University, 2020) Makiya, David NyangauForex markets are full of uncertainties. The forces of demand and supply determine the price of the Kenyan shilling in the international market. The rates usually adjust depending on the prevailing status of the economy, politics and influences of the Central Bank of Kenya (CBK) policies. Forex dealers are the dominant operators within the Kenyan forex market with forex bureaus and commercial banks taking the lead amongst them. A bureau would have a different pricing of a currency against the shilling but would nonetheless be within the bid-ask (buy sell) spread of the CBK ratings. Various online forex trading platforms have been implemented to facilitate trade in the Kenyan market. Predicting forex market prices is quite complicated as a process and subjective in nature for forex dealers, economists and business persons. The potential to make loses due to poor speculative guesses is quite high for multinational organizations located in more than one economy. The aim of this study is to develop a model for forex market price prediction in the Kenyan market using the Central Bank of Kenya data. Using the Data-Driven modelling technique, a model for forex market price prediction has been developed based on historical data from the CBK. The dataset is divided into training and testing data by a splitting of 80-20 respectively. The unique behavior of each of the currency data necessitated separate implementation of the currencies on the model for increased accuracy and lower error levels hence efficiency and optimality. The prediction model is achieved by combining time series analytical techniques with resilient backpropagation neural network. Successful predictions are conducted of up-to eight months forward with accuracy levels ranging 88-98% and Sum of squared residual (SSE) of 0.496-2.667, hence showing that combining time series analytics to resilient backpropagation neural networks to create a forex market prediction model with unique implementation of each currency is optimal for forex market price prediction where more data depicts longer period predictions.
- ItemComparative sentiment analysis of techniques for cyberbullying detection on twitter(Strathmore University, 2020) Kanam, Victor OtienoCyberbullying has become a common vice on the social media platforms and is quickly running out of hand. The psychological researches conducted on its effect are showing dire trends on the victims, sometimes leading to suicides among the victims. Currently, the efforts by the social media sites in curbing cyberbullying is largely user centered. Twitter platform provides a series of reactionary measures of dealing with cyberbullying instances, including; blocking users, reporting users, deleting posts and tagging tweets with warning labels. However, these approaches are more of reactionary than preventive. This leaves a gap in the software systems design which should eliminate the human intervention, by implementing technological methods in curbing cyberbullying. This research implemented the application of machine learning techniques to build a text classifier to detect instances of cyberbullying as the tweets are being composed. The research collected data from Twitter which was processed and labelled appropriately. A Support Vector Machine model was developed, trained and validated based on labelled text data using bigram features and term frequency-inverse document frequency weighting. An experimental approach was taken in determining what combination of features provided the most desirable performance outcome on the data collected. A comparative analysis was then done between the text classification algorithms (including Naïve Bayes, K-Nearest Neighbor and Random Forest Classifier) coupled the different features. The SVM classifier coupled with the bi-gram feature emerged as the best classifier while using sentiment to classify texts documents, with an accuracy of 84.22%.
- ItemDetecting financial crimes using pattern recognition techniques: case of mobile money transactions(Strathmore University, 2020) Eshiwani, Michelle MercyFinancial Crimes have evolved and gained complexity in the recent past owing to advanced technological adoption globally. As consumers have accepted new forms of service delivery that offer them convenience, affordability and easy access, criminals have also found new avenues of pushing their illegal funds or financing criminal activities without raising suspicion or being detected. It is therefore widely recognized that the prevalence of economically motivated crime in many societies is a fundamental threat to the development of world economies and their stability. This research aimed to develop a pattern recognition tool to analyze transaction patterns and detect suspicious transactions. This would in turn reduce the impact of financial crimes on mobile money transactions in terms of loss of revenue for both individuals, corporations and countries by safeguarding legitimate transactions while also tying any loose ends that facilitate the transfer of illegally acquired funds over legitimate channels. This research focused on the field of Pattern Recognition in identifying and analyzing fraud in mobile money transactions. The tool applied Statistical Pattern recognition using the K-Nearest Neighbor algorithm to accurately classify transactions as fraudulent or genuine.
- ItemA tool for mapping and monitoring landslides emergency management and disaster response: case study Murang’a County(Strathmore University, 2020) Kimani, Michael NgugiMurang’a County is considered the county which is more prone to landslides than any other part in Kenya. This is mainly due its mountainous terrain that has rugged a landscape composed of steep valleys punctuated by numerous hills. The terrain is dissected therefore creating the menace of landslides that come often during the rainy season. According to Kenya Red Cross, they have reported that Murang’a County has recorded the highest number of loss of life as well as property destruction as a result of landslides. The magnitude of the landslide often stretch to about three kilometres making it difficult to reach the affected villages. There has been a big challenge in identifying the impact of the landslides and infrastructures affected, hampering the coordination of emergency response efforts mainly because the data is not integrated spatially. Most of infrastructure damaged are people’s houses, roads, tea factories, tea buying centres, schools, hospitals, the tea farms not mention loss of human life and animals. To address this challenge, a tool that utilizes location intelligence as a spatial analytic technique to map and monitor landslide emergencies as well as respond to disasters in a more informed manner was developed. Spatial analysis lends new perspectives to a decision-maker as they study landslide occurrence, households destroyed, infrastructure affected and the relationships among them in an easily understandable manner. The tool was used to record & monitor landslide events, using an interactive operation dashboard that spatially showed where the landslides occurred, location of affected households and damaged infrastructure so as to coordinate the response services deployed. The tool was anchored on location intelligence, a spatial analysis technique, which provided various ways of analysing landslides events geographically and integrated infrastructure data to determine the likely impact. The findings of the research showed that users found the application informative and easy to use. The users were able to locate the areas where landslides often occur and were satisfied with the useful information that assisted them in identifying the infrastructure that was at risk.
- ItemBlock Chain Technology to enhance food traceability and safety: case study of Agriculture Industry in Kenya.(Strathmore University, 2020) Lesiit, Lenjula LetitoiyaFood safety has been one of the growing concerns and challenges in African countries. We had reports of sub standards food stuff, including chemically made eggs being sold to un-suspecting citizens. This is a global challenge, however, it is worth mentioning that it has been a serious issue in Africa due to rampant corruption where the oversight institution are easily compromised by unscrupulous business entities mainly because of lack of tracking system that enable the public know who made decisions or certain approvals. In South Africa, Listeria Outbreak caused 203 fatalities by the time the outbreak was declared over in September 2018. Closer home, Kenya has seen a fair share of food scandal starting with the infamous Aflatoxin-contaminated maize in Kenya resulted in 317 cases of hepatic failure and 125 deaths. There is lack of a system that would ensure companies and individuals involved in production of agricultural produce keep highest level of ethics; through a transparent supply chain management system that not only give the policy makers and oversight organization openness but also neutrality, reliability and thus security of the produce from the farm to table. This projects main objective is to develop a system that will guarantee food quality and safety from supply chain perspective by applying blockchain technology in agricultural supply-chain management, from farm to table. Since its conception in 2008, blockchain has developed over the last decade into one of today’s biggest technologies with a massive potential to impact virtually every industry from financial to manufacturing to educational institutions. Blockchain provided the answer to digital trust because it records important information in a public space and doesn’t allow anyone to remove it; it’s transparent, time-stamped and decentralized. The system developed enhances food safety and integrity through higher traceability thus helping everybody stakeholder in the supply chain quickly trace outbreaks back to specific sources, which could mitigate food fraud or food crises. The stakeholder are not only able to get quality and safer products but also forces the dishonest business entities out of business therefore making the market safe. The system offers many other benefits tandem to the blockchain attributes such as, providing a secure way to perform transactions among untrusted parties; decentralizing ledger that helps in connecting inputs, suppliers, producers and buyers.
- ItemApplication of permissioned block chain technology on population data consolidation and sharing(Strathmore University, 2020) Omoka, Richard Siang'aniPopulation registers should provide the single source of truth for data regarding each resident of a jurisdiction of the register, over the lifetime of the individual. This data can then be shared and used by government agencies and private organizations regarding matters concerning the individual. In Kenya, however, data regarding an individual is collected by multiple government agencies resulting in duplication (of effort and data) and data inconsistency. The multiple collections of population data result in an individual having multiple valid identification documents. The use of relational database management systems, which have shortcomings in support for temporal data as well as no inbuilt security and auditability capability makes relational database management systems ineffective in the storage of population data. Lack of clear policy and standards; interoperability issues and data security are among the challenges affecting data sharing among government agencies. Blockchain technology, a shared, immutable, distributed ledger that facilitates the process of recording transactions and tracking assets in a business network, is a promising technology in the management of population registers. Blockchain technology has inbuilt capacity to solve most of the problems inherent in the current systems especially duplication, tampering, and sharing of data. This research, therefore, through the development of a prototype based on permissioned blockchain technology, explores the viability and validity of permissioned blockchain technology, in storing, securing, auditing, and sharing of population data to achieve the single source of truth of the population register. The prototype, implemented using a local installation of hyperledger fabric, enabled consolidation of data since all invited participants on the permissioned blockchain network were able to write data to the single blockchain. The invited participants were also able to read data off the chain based on defined access control rules therefore achieving a uniform standard for data sharing. Provenance, a key quality of blockchain was leveraged to track an individual’s data changes over time, with the current block holding the latest records about the individual, yet still maintaining the historical chain of an individual’s data changes. This was a key outcome especially because it solves the inability of relational database systems to support temporal data. This model for data consolidation and sharing was found to be simple in design and implementation since it provided a standard way of reading and writing data to the chain through the use of RESTful APIs.
- ItemPredicting breast cancer progression by using cell-free DNA(Strathmore University, 2020) Bwire, AlbertCancer is among the leading causes of deaths in Kenya after infectious and cardiovascular diseases. Among the various forms of cancer, breast cancer accounts for a significant percentage of all new cancer incidences in the country and has a high mortality rate. On a global level, breast cancer is considered the most common cancer. Treatment methods employed vary from patient to patient due to factors such as the stage, age, and health. Treatment methods such as surgery, radiotherapy, chemotherapy or a combination of all have been used all to varying degrees of success and are not always efficient. However, these modalities have been employed successfully when the disease is detected early. This research applied deep neural networks coupled with genetic algorithms to build a learning model that evaluated the biomarkers obtained from cell-free DNA. The model was able to predict progression of breast cancer. The research, in addition, employed an agile, data-driven methodology due to its recursive nature producing a model with a higher degree of accuracy and specificity. The model developed was able to attain an accuracy of 94% in predicting breast cancer progression.
- ItemA secure hybrid IoT architecture using blockchain and decentralized VPN: a use case of smart farming(Strathmore University, 2020) Hilim, Peleke madinaThe Internet of Things (IoT) technology has brought a revolution to every sector of man's life by making it interactive, insightful, and smart. IoT refers to a network of objects that make up a network that configures itself. The proliferation of smart farming IoT-based devices transforms the face of farming turn over day after day by not just only boosting it but also making it cost-effective and reducing wastage. Nonetheless, many IoT characteristics can lead to matters of security and privacy, such as connectivity, wireless, embedded usage, variety, and scale. These features make IoT special in its security requirements and pose numerous new information security threats. The objective of this study is to propose a secure IoT based smart farming using Blockchain Technology and decentralized VPN for secure and efficient environment monitoring which could encourage farmers and other stakeholders, despite other challenges, to embrace smart farming and increase their overall yield and quality of products. In this paper, the proposed integrated solution is based on Hyperledger and OpenVPN technology. The architecture is based on a wireless mesh network with robust encryption at the network and application layer. The proposed solution is resilient to the common wireless network attack.
- ItemAwaking guarantee in mobile wireless networks for mobile nodes based on stochastic mobility mode(Strathmore University, 2020) Totona, Bruce LonyeiyeThe mobile wireless network is one of the areas in mobile computing getting more attention due to a plethora of innovations around the subject. The innovations and applications range from wireless sensor networks (Low-Rate Wireless Personal Area Network (LR-WPANs)), embedded systems and robotics. However, the technology provides some challenges in terms of mobile nodes awakening (active connection and synchronization) and services guarantee because the networks are highly dynamic. This research studied how to improve service provisions in mobile wireless networks and particularly in the wireless sensor networks using ZigBee modules by integrating synchronization and the routing procedure in the network. Ad-hoc On-demand Distance Vector (AODV) which is a common routing algorithm in wireless mobile networks was combined with the synchronization algorithm- non-beacon-enabled CSMA-based IEEE 802.15.4 MAC to overcome the challenge of hidden mobile nodes which are most often the new nodes joining the network. The beacon interval (Bi), the superframe duration (Sd), and the beacon time offset (Bto) were the key integration parameters and comes after active scanning. Random Waypoint (RWP) mobility model on Matlab was used to evaluate the mobility of the nodes considering their speeds, direction, and position from the coordinator. Elliptic Curve Digital Signature Algorithm (ECDSA) and the Advanced Encryption Standard (AES) algorithms were used to secure the modules and the channels.
- ItemA Representational state transfer web tool for firewall service management and monitoring in a Local Area Network(Strathmore University, 2020) Guchu, Mary WambuiFirewalls play a very important role in managing and securing organization resources. Firewalls implementation come in different designs and architectures. However, one thing that is common in most implementations is that they are complex to manage and configure. As the number of rules and policies on a firewall grow, rule proliferation, the complexity in managing these services grow. Firewall are mostly implemented using UNIX based systems. A number of them that range from IP tables and uncomplicated firewalls (UFW) are application that run host based firewalls system on a Linux environment. These services are mostly managed, configured and orchestrated using the command line. As the network grows and the need for complex rules arises, the management of such firewalls present a challenge. Developing web interfaces to monitor firewalls and network operations is one approach. However, a huge part of the challenge is in service deployment and orchestration of firewall services. This research proposes a system that system and network administrators can use to easily manage and configure firewalls without necessarily logging in to the console of the devices or machines or setting up secure shell (SSH) session to the command line interface (CLI). The web tool that this work proposes bases its argument on creating a web interface that allows system and network administrators to centrally manage firewall services on simple HTTP like interface with PHP REST APIs running in the background. This work used a test driven methodology coupled with agile development, specifically extreme programming to develop the RESTful web tool for easy firewall service management and monitoring.
- ItemA Block Chain Technology to enhance food traceability and safety: case study of agriculture industry in Kenya(Strathmore University, 2020) Lenjula, Letitoiya LesiitFood safety has been one of the growing concerns and challenges in African countries. We had reports of sub standards food stuff, including chemically made eggs being sold to un-suspecting citizens. This is a global challenge, however, it is worth mentioning that it has been a serious issue in Africa due to rampant corruption where the oversight institution are easily compromised by unscrupulous business entities mainly because of lack of tracking system that enable the public know who made decisions or certain approvals. In South Africa, Listeria Outbreak caused 203 fatalities by the time the outbreak was declared over in September 2018. Closer home, Kenya has seen a fair share of food scandal starting with the infamous Aflatoxin-contaminated maize in Kenya resulted in 317 cases of hepatic failure and 125 deaths. There is lack of a system that would ensure companies and individuals involved in production of agricultural produce keep highest level of ethics; through a transparent supply chain management system that not only give the policy makers and oversight organization openness but also neutrality, reliability and thus security of the produce from the farm to table. This projects main objective is to develop a system that will guarantee food quality and safety from supply chain perspective by applying block chain technology in agricultural supply-chain management, from farm to table. Since its conception in 2008, block chain has developed over the last decade into one of today’s biggest technologies with a massive potential to impact virtually every industry from financial to manufacturing to educational institutions. Block chain provided the answer to digital trust because it records important information in a public space and doesn’t allow anyone to remove it; it’s transparent, time-stamped and decentralized. The system developed enhances food safety and integrity through higher traceability thus helping everybody stakeholder in the supply chain quickly trace outbreaks back to specific sources, which could mitigate food fraud or food crises. The stakeholder are not only able to get quality and safer products but also forces the dishonest business entities out of business therefore making the market safe. The system offers many other benefits tandem to the block chain attributes such as, providing a secure way to perform transactions among untrusted parties; decentralizing ledger that helps in connecting inputs, suppliers, producers and buyers.
- ItemPrediction model for determining healthcare facility locations by the Kenyan County Government(Strathmore University, 2020-03) Isaboke, Edward NyakerumaHealth has been a key subject of interest to governments and non-governmental organizations. Health includes various building blocks with human resource being the core of health systems. Evidence globally shows a correlation between a country’s healthcare work force and health. Infrastructure relating to health is crucial for efficient and effective healthcare systems. A strategic location for a healthcare facility can improve facility utilization and save costs. Knowledge of methods and techniques need to be regularly updated for the location and establishment of healthcare facilities. This research investigates the criteria for decision-making of locations of healthcare facilities to ensure the most strategic location for the facilities is opted for. This research looks at the challenges that affect the location of healthcare facilities in detail to model the most appropriate locations for the facilities. Existing location models developed are many and discussed in detail to cater for special cases. The two most important criteria for healthcare services are cost and efficiency. They enable reduced distance to travel by patients to the health facilities. Unpredictability is identified as an unavoidable element of healthcare facilities location problem. The purpose of this research is to develop a model for determining strategic locations to establish healthcare facilities. The model was developed using the rapid application development methodology. This research proposed a strategy which enabled predictions according to the location factors that are suitable for establishment of healthcare facilities. This is after a set of locations are identified to be the target locations by the set cover location model. The model calculates and identifies the strategic locations within an area entered by a user. It is therefore important for historical data to exist for the model to work. Based on the provided data, the model then calculates the priority of the strategic locations identified hence a sequence is known on which location to establish the facility. Datasets from an online source were used as data inputs into the prediction model. The research recommended the model as a base for decision making for establishment of a healthcare facilities.
- ItemA Public complaints data mining and visualization tool for social media: a case of Nairobi City County(Strathmore University, 2020-04) Olweru, StephanieThe government of Kenya has been experiencing low levels of public complaint feedback with regards to the current complaint reporting methods. Citizens consider the organized public meetings and office visits as time consuming and opt out of the complaint reporting process. However, citizens have been taken to complaining on social media to express their dissatisfaction with infrastructural service delivery by government. Recent efforts to improve low levels of interaction between governments and their citizens by mining data from social media have proven that citizen sourcing from social media is a suitable approach to solving this issue. Social media posts, comments, likes, favourites, shares, retweets and similar features are clear indicators of public feedback if the posts are related to public service. This study proposed the development of a tool that extracts and visualizes public complaints relating to Nairobi county from social media. Roads were the focus infrastructure category for the study. For classification of collected tweets, a Support Vector Classifier (SVC) with TF-IDF features was trained and tested using labelled tweets. The classifier was able to label tweets collected with an accuracy of 77.52%. Location information was retrieved from the tweets classified as complaints using the Stanford Core NLP named entity recognizer. The locations identified in the complaint tweets using the Stanford NER were reverse geocoded using the Google Geocoding API and the resulting geographic coordinates plotted on a static google map. Frequent words in the complaint tweets were represented using a WordCloud. The county government can use this information for decision making, planning and to give feedback to the public on resolution of the same.
- ItemA Computer vision based model for tomato plant nutrient and disease classification(Strathmore University, 2020-06) Kiyegga, Raymond PaulDetermination of disease and nutrients in plants is still a new concept. Despite efforts from researchers to come up with improved techniques of detecting diseases and nutrients, many have been limited to only specific plant images and no other data such as weather, surrounding conditions to back up the decision. Plant disease identification is very crucial to food production and security, however current practices in Africa include visual identification and microscopy. Visual methods are greatly affected by cognitive error while microscopy is time consuming. It is difficult to detect plant disease unless one is guided by expert knowledge. Therefore, there is a need to apply machine learning techniques to make use of this expert knowledge. Current practices include use of spectral images to achieve this in fruits and other applications to help farmers without access to this knowledge to diagnose plant diseases. One notable challenge is determining nutrient content using images. Current applications require a farmer to look at the provided image and compare with what he sees on the plant. This research work proposes a machine learning model that can automatically detect the disease affecting a tomato plant as well as the nutrition level in the plant leaves. The farmer captures an image on their phone while in the plantation, based on the features from the leaf, the model analyses the image and returns the details of the classification in terms of type of disease and presence of deficiency of nutrients. The model was built on convolution neural network and achieved an accuracy of 85% using a learning rate of 0.001. It trained on 8000 samples using 30 epochs. The model was trained, validated and tested.
- ItemVegetation index based crop yield prediction model using convolution neural network - a case study of Kenya(Strathmore University, 2020-06) Chepngetich, JudithPredicting the crop yield is a vital food security strategy that can help a country take suitable measures and come up with policies that will help in crop production management. Such predictions will also support the farmers and industries involved in crop production for strategizing the logistics of their business or farming activities. Having sufficient production plans can improve food sufficiency and avoid situations of food emergencies. Climate change has had a huge impact on food production with variations in crop yields, creating uncertainty. Most of the studies on crop yield prediction have been done based solely on weather data, which is sometimes inaccurate due to scarcity of weather information especially in developing countries, where there is poor record keeping and insufficient resources to collect data. The use of vegetation indices derived from remote sensing data overcomes these challenges by providing data that is easily accessible and gives a comprehensive and multidimensional analysis. This study proposes a model that uses of vegetation index to predict crop yield using machine learning. Data from past crop yields in Kenya and vegetation greenness indices were the inputs applied to the algorithms. Various machine learning algorithms were applied and thereafter evaluated, so as to select the algorithm that gives better accuracy. To determine the accuracy level for the prediction model, the RMSE is calculated to compare actual and predicted values. The RMSE values obtained using convolution neural network for the three crops maize, rice and wheat were lower compared to those obtained using ridge regression, so it was selected as the optimal algorithm for the crop yield prediction model.