NRU :: Browsing by Author "Gidudu, Anthony"

Browsing by Author "Gidudu, Anthony"

Now showing 1 - 20 of 27

Assessment of irrigation water distribution using remotely sensed indicators: A case study of Doho Rice Irrigation Scheme, Uganda
(Smart Agricultural Technology, 2023) Wamala, Fawaz; Gidudu, Anthony; Wanyama, Joshua; Nakawuka, Prossie; Bwambale, Erion; Chukalla, Abebe D.
The rising competition for scarce land and water resources and the need to satisfy the global food demand from an ever-growing population necessitates novel methods to monitor irrigation scheme performance for improved water use efficiency. The traditional methods employed in sub-Saharan Africa to assess irrigation performance are point-based, expensive, and time-consuming, making monitoring and evaluation of these capital-intensive projects difficult. This study aimed at employing satellite data with high spatial and temporal resolution in assessing the performance of Doho Rice Irrigation Scheme through estimations of actual evapotranspiration. Actual evapotranspiration (ETa) was modelled from Landsat 7 imagery using the surface energy balance system algorithm on five clear days between January and April 2020. Using equity and adequacy metrics, the derived ETa was used to assess the irrigation performance of the scheme. Results showed that the equity indicator was generally fair, with the coefficient of variation between 0.11 and 0.08, close to the 0.10 threshold implying irrigation water is fairly distributed within the scheme. The average adequacy was 0.87, above the 0.65 threshold, indicating adequate water supply throughout the scheme. The study’s findings can be used in future research and benchmarking with other irrigation schemes to address the country’s water resource management challenges.
Classification of Images Using Support Vector Machines
(arXiv preprint arXiv, 2007) Gidudu, Anthony; Hulley, Greg; Tshilidzi, Marwala
Support Vector Machines (SVMs) are a relatively new supervised classification technique to the land cover mapping community. They have their roots in Statistical Learning Theory and have gained prominence because they are robust, accurate and are effective even when using a small training sample. By their nature SVMs are essentially binary classifiers, however, they can be adopted to handle the multiple classification tasks common in remote sensing studies. The two approaches commonly used are the One-Against-One (1A1) and One- Against-All (1AA) techniques. In this paper, these approaches are evaluated in as far as their impact and implication for land cover mapping. The main finding from this research is that whereas the 1AA technique is more predisposed to yielding unclassified and mixed pixels, the resulting classification accuracy is not significantly different from 1A1 approach. It is the authors conclusions that ultimately the choice of technique adopted boils down to personal preference and the uniqueness of the dataset at hand.
Comparison of Feature Selection Techniques for SVM Classification
(International Symposium on Physical Measurements and Signatures in Remote Sensing, 2007) Gidudu, Anthony; Ruther, Heinz
The use of satellite imagery in the derivation of land cover information has yielded immense dividends to numerous application fields such as environmental monitoring and modeling, map making and revision and urban studies. The extraction of this information from images is made possible by various classification algorithms each with different advantages and disadvantages. Support Vector machines (SVMs) are a new classifier with roots in statistical learning theory and their success in fields like machine vision have drawn the attention of the remote sensing community. Previous studies have focused on how SVMs compare with traditional classifiers such as maximum likelihood and minimum distance to means classifiers. They have also been compared to newer generation classifiers such as decision trees and artificial neural networks. In this research the understanding of the application of SVMs to image classification is furthered by proposing feature selection as a way in which remote sensing data can be optimized. Feature selection involves selecting a subset of features (e.g. bands) from the original set of bands that captures the relevant properties of the data to enable adequate classification. Two feature selection techniques are explored namely exhaustive search and population based incremental learning. Of critical importance to any feature selection technique is the choice of criterion function. In this research a new criterion function called Thornton’s separability index has been successfully deployed for the optimization of remote sensing data for SVM classification.
Computation of the Gravimetric Quasigeoid Model over Uganda Using the KTH Method
(Sofia Bulgaria, 2015) Ssengendo, Ronald; Sjöberg, Lars E.; Gidudu, Anthony
The gravimetric quasigeoid can be determined either directly by Stokes formula or indirectly by computing the geoid first and then determining the quasigeoid-to-geoid separation which is then used to determine the quasigeoid. This paper presents the computational results of the gravimetric quasigeoid model over Uganda (UGQ2014) based on the later technique. UGQ2014 was derived from the Uganda Gravimetric Geoid Model (UGG2014) which was computed by the technique of Least Squares Modification of Stokes formula with additive corrections commonly called the KTH Method. UGG2014 was derived from sparse terrestrial gravity data from the International Gravimetric Bureau, the 3 arc second SRTM ver4.1 Digital Elevation Model and the GOCE-only geopotential model GO_CONS_GCF_2_TIM_R5. The quasigeoid-to geoid separation was then computed from the Earth Gravitational Model 2008 (EGM08) complete to degree 2160 of spherical harmonics together with the global topographic model DTM2006.0 also complete to degree 2160. Another aim of this paper is to compare the approximate and strict formulas of computing the quasigeoid-to-geoid separation and evaluate their effects on the final quasigeoid model. Using 10 GNSS/levelling data points distributed over Uganda, the RMS fit of the quasigeoid model based on the approximate formula are 27 cm and 10 cm before and after a 4-parameter fit, respectively. Similarly, the RMS fit of the model based on the strict formula are 15 cm and 6 cm, respectively. The results show the improvement to the final quasigeoid model brought about by using the strict formula to model more effectively the terrain in the vicinity of the computation point. With an accuracy of 6 cm, UGQ2014 represents significant progress towards the computation of a final gravimetric quasigeoid over Uganda which can be used with GNSS/levelling. However, with more data especially terrestrial gravity data and GNSS/levelling we anticipate that the accuracy of gravimetric quasigeoid modelling will improve in future.
Determination of Satellite-Derived PM2.5 for Kampala District, Uganda
(Geomatics, 2022) Atuhaire, Christine; Gidudu, Anthony; Bainomugisha, Engineer; Mazimwe, Allan
Ground monitoring stations are widely used to monitor particulate matter (PM2.5). However, they are expensive to maintain and provide information localized to the stations, and hence are limited for large-scale use. Analysis of in situ PM2.5 shows that it varies spatially and temporally with distinct seasonal differences. This study, therefore, explored the use of satellite images (Sentinel-2 and Landsat-8) for determining the spatial and temporal variations in PM2.5 for Kampala District in Uganda. Firstly, satellite-derived aerosol optical depth (AOD) was computed using the Code for High Resolution Satellite mapping of optical Thickness and aNgstrom Exponent algorithm (CHRISTINE code). The derived AOD was then characterised with reference to meteorological factors and then correlated with in situ PM2.5 to determine satellite-derived PM2.5 using geographically weighted regression. In the results, correlating in situ PM2.5 and AOD revealed that the relationship is highly variable over time and thus needs to be modelled for each satellite’s overpass time, rather than having a generic model fitting, say, a season. The satellite-derived PM2.5 showed good model performance with coefficient of correlation (R2) values from 0.69 to 0.89. Furthermore, Sentinel-2 data produced better predictions, signifying that increasing the spatial resolution can improve satellite-derived PM2.5 estimations.
Effect of Landscape Changes on the Water Quality of Murchison Bay
(International Journal of Advanced Remote Sensing and GIS, 2020) Sridhar, Balaji Bhaskar Maruthi; Gidudu, Anthony
The water quality in Murchison Bay of Lake Victoria, the Africa’s largest fresh water lake, is on decline due to rapid urban sprawl, decrease in vegetative surface and increase in impervious surface of the drainage area resulting in eutrophication of the lake. The objectives of our study are 1) to analyze the nutrient and metal concentrations in the Murchison Bay; 2) to identify and map the long-term landscape changes in Murchison Bay Watershed (MBW); 3) to analyze the impact of the landscape changes on the water quality of Murchison Bay. Water samples were collected from Miami Beach (MB), Ggaba Beach (GB) and Mulungo Beach (MuB), along the Murchison Bay and analyzed for various metal and nutrient concentrations. Landsat satellite imagery, sampled over three decades (1995-2019) of the MBW were analyzed for the land cover changes. The chemical analysis of the water samples showed that the P concentrations were above the critical limits while the As and Pb concentrations were higher but remained below the critical limits in water. The remote sensing analysis reveal that the impervious surface in the MBW increased by about 21.9% while the vegetative surface decreased by 4.2% during the period of 1995 to 2019. The Chlorophyll a concentration in the Murchison Bay increased over the period of time resulting in deterioration of water quality. Integration of environmental chemical analysis along with geospatial data aids in understanding the impact of land scape changes on the Murchison Bay water quality and to identify the areas vulnerable to change.
An Empirical Evaluation of Data Interoperability—A Case of the Disaster Management Sector in Uganda
(ISPRS International Journal of Geo-Information, 2019) Mazimwe, Allan; Hammouda, Imed; Gidudu, Anthony
One of the grand challenges of disaster management is for stakeholders to be able to discover, access, integrate and analyze task-appropriate data together with their associated algorithms and work-flows. Even with a growing number of initiatives to publish data in the disaster management sector using open principles, integration and reuse are still difficult due to existing interoperability barriers within datasets. Several frameworks for assessing data interoperability exist but do not generate best practice solutions to existing barriers based on the assessment they use. In this study, we assess interoperability for datasets in the disaster management sector in Uganda and identify generic solutions to interoperability challenges in the context of disaster management. Semi-structured interviews and focus group discussions were used to collect qualitative data from sector stakeholders in Uganda. Data interoperability was measured to provide an understanding of interoperability in the sector. Interoperability maturity is measured using qualitative methods, while data compatibility metrics are computed from identifiers in the RDF-triple model. Results indicate high syntactic and technical interoperability maturity for data in the sector. On the contrary, there exists considerable semantic and legal interoperability barriers that hinder data integration and reuse in the sector. A mapping of the interoperability challenges in the disaster management sector to solutions reveals a potential to reuse established patterns for managing data interoperability. These include; the federated pattern, linked data patterns, broadcast pattern, rights and policy harmonization patterns, dissemination and awareness pattern, ontology design patterns among others. Thus a systematic approach to combining patterns is critical to managing data interoperability barriers among actors in the disaster management ecosystem.
An Empirical Investigation of GIS Interoperability Best Practices In Industry
(Preprints, 2018) Mazimwe, Allan; Hammouda, Imed; Gidudu, Anthony
Reuse of patterns is a self-evident approach for managing interoperability concerns. Although patterns for resolving interoperability barriers exist in the literature, no study exists on adoption of interoperability patterns by Geographic Information Systems (GIS) practitioners in industry. Thus there is limited understanding of pattern re-usability, yet the advantages offered by interoperability patterns provide a reasonably sound justification for their usage. This paper examines the adoption of proven interoperability best practices in the GIS industry. An empirical study that involved the use of semi-structured interviews was employed to gather data from GIS developers on domain interoperability best practices. Results indicated that industry and communities of practice have been converging on the technical level to ensure interoperability of GIS concerns. Semantic interoperability and related patterns are least understood, yet semantic barriers still exist. This is partly due to the complexity associated with the top-down approach used to develop semantic interoperability solutions. Therefore, this study proposes research into resolving barriers in the adoption of interoperability patterns that reduce complexity while solving semantic interoperability barriers.
Empirical modeling of chlorophyll a from MODIS satellite imagery for trophic status monitoring of Lake Victoria in East Africa
(Journal of Great Lakes Research, 2021) Gidudu, Anthony; Letaru, Lydia; Kulabako, Robinah N.
We detail our attempts at empirical modeling of MODIS derived Chlorophyll a (Chl a) distribution on Lake Victoria in East Africa and consequently its trophic status. This was motivated by the need for Lake Victoria specific algorithms, as the current satellite based standard algorithms overestimate derived Chl a. In situ Chl a data was hence collected in three field campaigns in November 2014, March 2015 and July 2015. In situ reflectances were collected during the July campaign only. We first developed models from in situ reflectances and in situ Chl a, which when applied to MODIS bands performed dismally (R2 = 0.03). We then proceeded to derive empirical models by directly comparing MODIS bands with in situ Chl a based on data collected in November 2014 and July 2015. The March 2015 dataset couldn’t be used due to cloud cover hence no matchups could be obtained. The best model derived (R2 = 0.88) was based on the ratio 488 nm/645 nm, and was then used to determine the trophic status of Lake Victoria using Carlson’s Chl a Trophic State Index (TSI). The results show that large areas of the lake are mesotrophic with eutrophic displays closer to the shores. The modeled TSI was then validated against in situ TSI derived from the March dataset and posted an 80% matchup. One of the main challenges, however is the prevalence of cloud cover, which hinders synoptic mapping of the lake. That notwithstanding, the study demonstrates the potential of earth observation in providing accurate TSI information for improved management of Lake Victoria.
Exploring MODIS Imagery in Monitoring Water Quality on Lake Victoria
(Journal of Great Lakes Research, 2021) Gidudu, Anthony; Mpakiraba, Zainab; Kalibbala, Herbert
At 68,800 km2, Lake Victoria is the largest fresh water lake in Africa. It is a trans-boundary water resource supporting the livelihoods of over 20 million people directly and indirectly. It is a source of food, recreation, domestic and industry use. This has rendered its monitoring of paramount interest to several environmental agencies in Uganda, Kenya and Tanzania as well as along the river Nile basin. Traditionally, the monitoring of water quality is carried out at specific points of the lake by carrying out in-situ measurements or collection of water samples for laboratory testing. This traditional approach of determining water quality is cumbersome, expensive and does not give a synoptic perspective of the whole lake. This has inspired the consideration of satellite imagery as a tool to monitor water quality on the lake. Satellite imagery offers the advantage of providing regularly collected data, giving a synoptic view of the water quality of the whole lake. The aim of this paper was to therefore investigate how satellite derived water quality parameters compare with in situ measurements in a bid to operationalise the use of satellite images in monitoring water quality on the lake. To wit, in-situ lake surface temperature at specific points was measured and water samples of those points were taken to the lab to test for Chlorophyll_a. These samples were collected within 4 hours of satellite overpass. These results were then compared with water quality parameters derived from MODIS imagery. The results showed that there is a moderate to strong correlation (R2 = 0.68) between satellite derived lake surface temperature and in-situ measurements implying that MODIS satellite imagery can be depended to accurately model the spatial variation of lake surface temperature. Unfortunately because of cloud cover coinciding with the day of in-situ observations, no similar comparisons could be made regarding Chlorophyll_a, thus portraying one of the challenges of operationalizing the use of satellite imagery in monitoring water quality on lake Victoria. Given the potential of satellite imagery as a reliable source of water quality information, further studies are urgently needed to validate it for Lake Victoria.
A framework for the geometric accuracy assessment of classified objects
(International Journal of Remote Sensing,, 2013) Möller, Markus; Birger, Jens; Gidudu, Anthony; Gläßer, Cornelia
European initiatives for data harmonization and the establishment of remote-sensingbased services aim at the production of up-to-date land-cover information according to generally valid standards for the accurate qualification of thematic classification results. This is particularly true since new satellite systems provide data of high temporal and geometric resolution. While methods for point-related thematic accuracy assessment have already been established for years, there is a need for a commonly accepted framework for the geometric quality of tematic maps. In this study, an open and extendable framework for the geometric accuracy assessment is presented. The workflow begins with the definition of basic geometric accuracy metrics, which are based on differences in area and position between samples of classified and reference objects. The combination of user-defined metrics enables both a geometric assessment of single objects as well as the total data set. In an example of thematically classified agricultural fields in a German test site, we finally show how object relations between classified and reference objects can be identified and how they affect the global accuracy assessment of the total data set.
Geoid Determination In Uganda: Current Status
(Advances in Engineering and Technology, 2017) Ssengendo, Ronald; Sjöberg, Lars.E.; Gidudu, Anthony
Many professionals e.g. surveyors, engineers and GIS specialists are increasingly using Global Positioning System (GPS) or some other Global Navigation Satellite Systems (GNSS) for positioning and navigation. One of the greatest advantages of GPS is its ability to provide three-dimensional coordinates (latitude, longitude and height) anywhere in the world, any time irrespective of the weather. The GPS latitude and longitude can easily be transformed from the WGS84 reference system to a local reference (e.g. Arc 1960). However the GPSdetermined heights, i.e. ellipsoidal heights, are geometrical heights which have no physical meaning and therefore cannot be used in surveying and engineering projects. Their conversion to more meaningful orthometric heights require knowledge of the geoidal undulations, which can be determined from high resolution geoid models. Its absence in Uganda means that the full potential of GPS cannot be fully realized. This paper gives an overview of the need for an accurate geoid model in Uganda, the current status of the geodetic network in Uganda and different methods of geoid determination. Pending further investigation, preliminary findings indicate that in Uganda, the EGM2008 is the best geoid modal for GPS/leveling projects.
Implementation of FAIR Principles for Ontologies in the Disaster Domain: A Systematic Literature Review
(International Journal o f Geo-Information, 2021) Mazimwe, Allan; Hammouda, Imed; Gidudu, Anthony
The success of disaster management efforts demands meaningful integration of data that is geographically dispersed and owned by stakeholders in various sectors. However, the difficulty in finding, accessing and reusing interoperable vocabularies to organise disaster management data creates a challenge for collaboration among stakeholders in the disaster management cycle on data integration tasks. Thus the need to implement FAIR principles that describe the desired features ontologies should possess to maximize sharing and reuse by humans and machines. In this review, we explore the extent to which sharing and reuse of disaster management knowledge in the domain is inline with FAIR recommendations. We achieve this through a systematic search and review of publications in the disaster management domain based on a predefined inclusion and exclusion criteria. We then extract social-technical features in selected studies and evaluate retrieved ontologies against the FAIR maturity model for semantic artefacts. Results reveal that low numbers of ontologies representing disaster management knowledge are resolvable via URIs. Moreover, 90.9% of URIs to the downloadable disaster management ontology artefacts do not conform to the principle of uniqueness and persistence. Also, only 1.4% of all retrieved ontologies are published in semantic repositories and 84.1% are not published at all because there are no repositories dedicated to archiving disaster domain knowledge. Therefore, there exists a very low level of Findability (1.8%) or Accessibility (5.8%), while Interoperability and Reusability are moderate (49.1% and 30.2 % respectively). The low adherence of disaster vocabularies to FAIR Principles poses a challenge to disaster data integration tasks because of the limited ability to reuse previous knowledge during disaster management phases. By using FAIR indicators to evaluate the maturity in sharing, discovery and integration of disaster management ontologies, we reveal potential research opportunities for managing reusable and evolving knowledge in the disaster community.
Increased-specificity famine prediction using satellite observation data
(First ACM Symposium on Computing for Development, 2010) Quinn, John A.; Okori, Washington; Gidudu, Anthony
This paper examines the use of remote sensing satellite data to predict food shortages among di erent categories of house- holds in famine-prone areas. Normalized Di erence Vegeta- tion Index (NDVI) and rainfall estimate data, which can be derived from multi-spectral satellite radiometer images, has long been used to predict crop yields and hence famine. This gives an overall prediction of food insecurity in an area, though in a heterogeneous population it does not directly predict which sectors of society or households are most at risk. In this work we use information on 3094 households across Uganda collected between 2004-2005. We describe a method for clustering households in such a way that the cluster de- cision boundaries are both relevant for improved-speci city famine prediction and are easily communicated. We then give classi cation results for predicting food security status at a household level given di erent combinations of satel- lite data, demographic data, and household category indices found by our clustering method. The food security classi - cation performance of this model demonstrates the potential of this approach for making predictions of famine for speci c areas and demographic groups.
Introduction to the AARSE2016 Special Issue of the South African Journal of Geomatics
(South African Journal of Geomatics, 2018) Gidudu, Anthony; Bamutaze, Yazidhi
The papers in this special issue of SAJG are derived from the 11th International Conference of the African Association of Remote Sensing of the Environment hosted in Uganda by Makerere University. The conference was held from the 24th to 28th October 2016 under the theme ‘Our Earth, Our Heritage: Harnessing Geospatial Technologies for Sustainable Development in Africa’, with a view of recognizing not only the past efforts related to resource management challenges, but also positioning to take care of the expected changes, owing to the post 2015 Global Development agenda. The conference attracted scholarship under the following sub themes: Space and earth observation technology for sustainable development goals (SDG’s), Disaster risk management and resilience, From climate change to climate risk management, Human capital development in geospatial science, Big Data and spatial data infrastructure utility and management, Africa space policy and strategy: Cost benefit of space technology, Geospatial science and technology for water and watershed management, Conflict management, human security and peace, Cities and demographic transitions, Geospatial technologies for Energy Management. From these presentations, selected high-quality papers were invited to develop and submit full papers to the SAJG for consideration in a special issue. These papers were subjected to the journal’s review process which involved a double blind peer-review and accepted papers are included in this special issue on successful address of raised concerns.
An Investigation of the Relationship between Standard Penetration Test and Shear Wave Velocity for Unsaturated Soils (A Case Study of The Earthquake Prone Area of the Albertine Graben)
(Earthquake Engineering and Seismology Istanbul, 2014) Tumwesige, Robert; Gidudu, Anthony; Bagampadde, Umaru; Ryan, Conor
When an earthquake occurs, seismic waves radiate away from the source and travel rapidly through the earth’s crust. When these waves reach the ground surface, they cause shaking that may last from a few seconds to minutes. The nature and distribution of earthquake damage is strongly influenced by the response of soils to dynamic loading. This response is controlled to a large extent by the dynamic soil properties such as stiffness, damping, Poisson’s ratio and density. The prediction of ground shaking at soil sites requires knowledge of the soil expressed in terms of shear wave velocity (Vs). It is preferable to measure Vs by in situ wave propagation tests. However, it is often not economically feasible to conduct these tests at all locations. On the other hand the Standard Penetration Test (SPT) is the most common in situ site geotechnical test which is carried out in most site investigations. Hence a reliable correlation between Vs and SPT would be of considerable advantage, reducing the cost of site investigations. This paper presents, therefore, the development of an empirical relationship between Vs and SPT N-value for the soils of Kasemene Oil exploration area located in Buliisa District in Uganda. As part of an attempt to mitigate the effects of earthquakes in the area, a model is needed to predict Vs required for site response analysis. The effect of correcting Vs and SPT N-values on the model was evaluated and the model was also compared with published models. The process involved correlating 273 data pairs of Vs and SPT N-values which were measured at the same depth. The extensive Vs measurement was carried out using the Multichannel Analysis of Surface Waves (MASW) technique. The SPT N-value data was measured from boreholes drilled within the boundaries of the MASW survey lines. Results show that the relationship between Vs and SPT N-value depends on the effective overburden stress, and that ignoring the influence of effective overburden stress creates bias in the model. It was also found out that none of the published models fitted the data well and there is tremendous difference in the Vs values predicted by these models. The model exhibits good prediction performance and can be used to predict shear wave velocity for soils within the study area or for areas with a similar soil type.
Lockdown lessons: an international conversation on resilient GI science teaching
(Journal of Geography in Higher Education, 2022) Blanford, Justine I.; Bowlick, Forrest; Gidudu, Anthony; Gould, Michael; Griffin, Amy L.; Kar, Bandana; Kemp, Karen; de Róiste, Mairéad; de Sabbata, Stefano; Sinton, Diana; Strobl, Josef; Tate, Nicholas; Toppen, Fred; Unwin, David
We report the findings from two global panel “conversations” that, stimulated by the exceptional coronavirus pandemic of 2020/21, explored the concept of resilience in geographic science teaching and learning. Characteristics of resilient teaching, both in general and with reference to GISc, are listed and shown to be essentially what might in the past have been called good teaching. Similarly, barriers to resilient teaching are explored and strategies for overcoming them listed. Perhaps the most important conclusion is a widespread desire not to “bounce back” to pre-COVID ways, but to use the opportunity to “bounce forward” towards better teaching and learning practices.
Ontology Design Patterns for Representing Knowledge in the Disaster Risk Domain
(IEEE, 2019) Mazimwe, Allan; Hammouda, Imed; Gidudu, Anthony
The success of disaster risk management efforts depend on the ability of multiple stakeholders to share disasterrelated information. Semantic integration of such heterogeneous information requires ontology building. The top-down-approach of ontology building has several disadvantages to knowledge representation. To support the process of ontology engineering, a bottom-up-approach that utilizes modular Ontology Design Patterns (ODPs) with weak dependencies can be used to overcome the disadvantages of the top-down-approach. To bridge the availability gap of patterns for representing disaster knowledge, the study identifies existing and emerging patterns that can be used to organize disaster knowledge. Based on the eXtreme Design (XD) methodology and key informant interviews, Competency Questions (CQs) were listed from domain stakeholders. Consequently, corresponding patterns covering the CQs were also identified and developed. This study identifies emerging patterns such as Event Type ODP for representing risky and hazardous events. The QualityCausation ODP is also identified for representing the causality nature of vulnerability. The resulting patterns are aligned to the DOLCE1 foundational ontology and can be used to organize data in the disaster domain.
A Pattern Driven Approach to Knowledge Representation in the Disaster Domain
(SN Computer Science, 2020) Mazimwe, Allan; Hammouda, Imed; Gidudu, Anthony; Barasa, Bernard
Access to integrated disaster-related data through querying is still a problem due to associated semantic barriers. The disaster domain largely relies on the top–down approach of ontology development. This limits reuse due to associated commitments and complex alignments within ontologies. Therefore, there is a need to utilize a bottom-up approach that reuses patterns for representing disaster knowledge. To bridge the availability gap of patterns for representing disaster knowledge, this study identifies existing and emerging patterns for reuse while organizing disaster data from multiple sector stakeholders. Based on the eXtreme Design (XD) methodology and key informant interviews, competency questions (CQs) were elicited from domain stakeholders. The CQs are matched with existing patterns from other contexts. Emerging patterns (e.g the Event Classification and Quality Dependence Description for Objects) are also developed for CQs not captured and subsequently tested using SPARQL queries characterising the CQs. It is in this context that this paper presents a characterisation of disaster risk knowledge using CQs and corresponding patterns (reusable and emerging) covering the knowledge. Accordingly, we illustrate a pattern-driven use case to organise drought hazard data for early warning purposes. This provides a powerful use case for adopting a pattern-based approach to knowledge representation in the disaster domain.
Random ensemble feature selection for land cover mapping
(IEEE, 2009) Gidudu, Anthony; Bolanle, Abe T.; Marwala, Tshilidzi
Random ensemble feature selection is a means through which diversity in ensemble systems is imposed by randomly selecting the features (bands) that constitute the base classifiers. This paper provides insight and discusses the interplay between the size of the resulting ensembles and the consequent classification accuracy. From the results, it was observed that classification accuracy increased more as the number of features per base classifier increases than as the number of base classifiers increases. That said however, classification accuracy was seen to increase with additional features up to a given limit beyond which increasing the number of features per base classifier did not significantly increase classification accuracy, a peaking effect probably due to Hughes phenomenon.

Browsing by Author "Gidudu, Anthony"

Results Per Page

Sort Options