Browsing by Author "Potgieter, Anet"

Now showing 1 - 11 of 11
  • Thumbnail Image
    Item
    Open Access
    An adaptive agent architecture for exogenous data sales forecasting
    (2006) Jedeikin, Jonathan; Potgieter, Anet; April, Kurt
    In a world of unpredictability and complexity, sales forecasting is becoming recognised as essential to operations planning in business and industry. With increased globalisation and higher competition, more products are being developed at more locations, but with shorter product lifecycles. As technology improves, more sophisticated sales forecasting systems are developed which require increasing complexity. We tum to adaptive agent architectures to consider an alternative approach for modelling complex sales forecasting systems. This research proposes modelling a sales forecasting system using an adaptive agent architecture. It additionally investigates the suitability of Bayesian networks as a sales forecasting technique. This is achieved through BaBe, an adaptive agent architecture which employs Bayesian networks as internal models. We develop a sales forecasting system for a meat wholesale company whose sales are largely affected by exogenous factors. The company's current sales forecasting approach is solely qualitative, and the nature of their sales is such that they would benefit from a reliable exogenous data sales forecasting system. We implement the system using BaBe, and incorporate a Bayesian network representing the causal relationships affecting sales. We introduce a learning adjustment component to adjust the estimated sales towards closer approximations. This is required as BaBe is currently unable to use continuous data, resulting in a loss of accuracy during discretisation. The learning adjustment additionally provides a feedback aspect, often found in adaptive agent architectures. The adjustment algorithm is based on the mean error calculation, commonly used as sales forecasting performance measures, but is extended to incorporate a number of exogenous variables. We test the system using the holdout procedure, with a 5-fold cross validation data-splitting approach, and contrast the accuracy of the estimated sales, provided by the system, with sales estimated using a regression approach. We additionally investigate the effectiveness of the learning adjustment component.
  • Thumbnail Image
    Item
    Open Access
    An assessment of the onset of summer rainy season in Southern Africa - case study of Botswana
    (2008) Cheruiyot, Denis C.; Potgieter, Anet
    The economies of most Sub-Saharan African countries are linked to the onset, reliability and performance of seasonal rainfall. Failure of seasonal rains may signal food deficits or worse. Farmers, water conservationists and government bodies responsible for food security, all have an interest in seasonal rainfall: onset, approximate dates for start of the season and probabilities for early, normal or late onset of rains. This knowledge enables them make crucial decisions as to the choice of crops, planting dates, management of dams, pasture and hydro-electric dams. In this thesis, daily rainfall data for 29 rainfall stations in Botswana for the years 1971 - 2004 was analyzed to determine Start-of-Season (505)/ Onset of summer rainfall. We used Principal Component Analysis to determine rainfall homogeneous zones in Botswana. Basically three regions were identified for October, November December (OND) rainfall months. Rainfall values in representative stations in each zone (Northern, Central and South-Eastern and Western regions) were correlated with Sea Surface Temperatures (SSTs) in global oceans to determine ocean regions that correlate well with Botswana rainfall. The onset dates were grouped into false, early, normal, late and failed onsets. Monthly rainfall and Rainfall Onsets for selected 14 rainfall stations and ten other weather parameters, (that include SSTs, Sea Level Pressures (SLPs) and climate indices) were placed in a spreadsheet. Emergent Situation Awareness (ESA) for dynamic Bayesian networks (DBN) was used to analyze this data. The ESA for DBN models temporal dependencies among the weather parameters and climate indices using Direct Acyclic Graphs (DAG). This innovative DBN technology, ESA, reveals more detailed information from complex models. It reveals what is currently happening over time in a domain of interest. Each of the parameters and climate indices revealed varying degrees of beliefs for early, normal, late or failed rainfall onsets in Botswana. Some of the parameters which showed higher degrees of beliefs are promising signals to the onset of summer rains.
  • Thumbnail Image
    Item
    Open Access
    Anomaly detection and prediction of human actions in a video surveillance environment
    (2007) Spasic, Nemanja; Potgieter, Anet
    World wide focus has over the years been shifting towards security issues, not in least due to recent world wide terrorist activities. Several researchers have proposed state of the art surveillance systems to help with some of the security issues with varying success. Recent studies have suggested that the ability of these surveillance systems to learn common environment behaviour patterns as well as to detect and predict unusual, or anomalous, activities based on those learnt patterns are possible improvements to those systems. I addition, some of these surveillance systems are still run by human operators, who are prone to mistakes and may need some help from the surveillance systems themselves in detection of anomalous activities. This dissertation attempts to address these suggestions by combining the fields of image understanding and artificial intelligence, specifically Bayesian Networks, to develop a prototype video surveillance system that can learn common environmental behaviour patterns, thus being able to detect and predict anomalous activity in the environment based on those learnt patterns. In addition, this dissertatio aims to show how the prototpe system can adapt to these anomalous behaviours and integrate them into its common patterns over a prolonged occurrence period.
  • Thumbnail Image
    Item
    Open Access
    Automated stock trading : a multi-agent, evolutionary approach
    (2008) Kruger, Kurt; Potgieter, Anet
    Stock market trading has garnered much interest over the past few decades as it has been made easier for the general public to trade. It is certainly an avenue for wealth growth, but like all risky undertakings, it must be understood for one to be consistently successful. There are, however, too many factors that influence it for one to make completely confident predictions. Automated computer trading has therefore been championed as a potential solution to this problem and is used in major brokerage houses world-wide. In fact, a third of all EU and US stock trades in 2006 were driven by computer algorithms. In this thesis we look at the challenges posed by the automatic generation of stock trading rules and portfolio management. We explore the viability of evolutionary algorithms, including genetic algorithms and genetic programming, for this problem and introduce an agent-based learning framework for individual and social intelligence that is applicable to general stock markets. Statistical tests were applied to determine whether or not there was a significant difference between the evolutionary trading approach and an accepted benchmark. It was found that while the evolutionary trading agents comfortably realised higher portfolio values than the ALSI, there was insufficient evidence to suggest that the agents outperformed the ALSI in terms of portfolio performance. Additionally, it was observed that while the traders combined knowledge from the expert traders to form complex trading models, these models did not result in any statistically significant positive returns. It must be said, however, that there was overwhelming evidence to suggest that the traders learned rules that were highly successful in predicting stock movement.
  • Thumbnail Image
    Item
    Open Access
    Bayesian networks for spatio-temporal integrated catchment assessment
    (2010) Dondo, C; Rivett, Ulrike; Chevallier, LP; Potgieter, Anet
    In this thesis, a methodology for integrated catchment water resources assessment using Bayesian Networks was developed. A custom made software application that combines Bayesian Networks with GIS was used to facilitate data pre-processing and spatial modelling. Dynamic Bayesian Networks were implemented in the software for time-series modelling.
  • Thumbnail Image
    Item
    Open Access
    Bayesian participatory-based decision analysis : an evolutionary, adaptive formalism for integrated analysis of complex challenges to social-ecological system sustainability
    (2010) Peter, Camaren; April, Kurt; Potgieter, Anet
    This dissertation responds to the need for integration between researchers and decision-makers who are dealing with complex social-ecological system sustainability and decision-making challenges. To this end, we propose a new approach, called Bayesian Participatory-based Decision Analysis (BPDA), which makes use of graphical causal maps and Bayesian networks to facilitate integration at the appropriate scales and levels of descriptions. The BPDA approach is not a predictive approach, but rather, caters for a wide range of future scenarios in anticipation of the need to adapt to unforeseeable changes as they occur. We argue that the graphical causal models and Bayesian networks constitute an evolutionary, adaptive formalism for integrating research and decision-making for sustainable development. The approach was implemented in a number of different interdisciplinary case studies that were concerned with social-ecological system scale challenges and problems, culminating in a study where the approach was implemented with decision-makers in Government. This dissertation introduces the BPDA approach, and shows how the approach helps identify critical cross-scale and cross-sector linkages and sensitivities, and addresses critical requirements for understanding system resilience and adaptive capacity.
  • Thumbnail Image
    Item
    Open Access
    Emergent Communication: The evolution of simplistic machines using different communication types
    (2009) Karpul, Alexander; Potgieter, Anet
    The methods of transmitting information may be divided as follows: direct; and, indirect. The âdirectâ method occurs when a creature transmits a signal that other creatures in its local environment can receive. Word of mouth advertising is a form of direct communication. âIndirectâ communication relays a message through the environment. This type of communication is known as stigmergy. Both word of mouth communication and stigmergy require the existence of groups of communicators. It is, however, difficult to analyse a very large number of local interactions that occur in group behaviour. A global phenomenon known as âemergenceâ arises from such behaviour. The phrase ââthe whole is greater than the sum of its partsâ normally describes emergence. In this research, we investigate how the two methods of communicating, direct and indirect (including a combination of these), result in emergent behaviour. In order to establish this outcome we employed the use of agent-based software in which we designed groups of agents to evolve over generations in response to specific situations. The manner in which these agent groups evolve is by a genetic algorithm. This is based on the consumption and collection of resources from the environment - a metric for gauging how well the population performs as a whole. For the purpose of this dissertation, we measure and examine the performance of four styles of the two methods of communication: No Communication, Word of Mouth, Stigmergic and Both (a combination of direct and indirect). We observe the fitness arising through successive generations of agents for each of the four styles and compare the results. The âNo Communicationâ style is markedly the worst performer and is âthe sum of the partsâ in terms of the definition of emergence. The âWord of Mouthâ style is marginally below the best performer but is rated well above that of âNo Communicationâ. The âStigmergicâ style is only the third best performer. Combining the direct and indirect methods yields the best result for the âBothâ style. All the communicating categories, considered âthe wholeâ in terms of the definition for emergence, outperform the âNo Communicationâ style. This demonstrates that emergence occurs when using these communication methods in groups. Keywords: Communication, Emergence, Genetic Algorithms, Group Behaviour
  • Thumbnail Image
    Item
    Open Access
    Estimating the spatial and temporal variability of primary production from a combination of in situ and remote sensing data a southern Benguela case study
    (2013) Williamson, Robert I; Field, John G; Shillington, Frank; Jarre, Astrid; Potgieter, Anet
    The aim of this thesis is to produce fine resolution estimates of primary production in three-dimensional space at the temporal scale that these events develop. It is hypothesized that complex relationships among time sequences of physical and biological processes that influence primary production can be automatically discovered from archives of data. This study uses an archive of in situ ship-board data containing subsurface temperature and phytoplankton distribution profiles. Each profile is associated in time and space with satellite remotely-sensed wind, sea surface temperature and surface chlorophyll a data. The bottom depth, season and location of each profile are also recorded. The archive of depth profiles is simplified by mapping each profile onto one of twelve representative profile clusters obtained using the k-means clustering algorithm so that each cluster contains a set of similar profiles and their corresponding data. Relationships between remotely sensed surface features and chlorophyll a profiles are first obtained from a static Bayesian network using same day data. This is then taken further by analysing time-series of satellite data to predict likely temperature and chlorophyll a profiles for each pixel of a 4 km resolution satellite image.
  • Thumbnail Image
    Item
    Open Access
    Intelligent detection of anomalies in telecommunications customer behaviour
    (2006) Osunmakinde, Isaac Olusegun; Potgieter, Anet
    In this research, we present a modelling technique that can efficiently facilitate anomaly detection that will help call analysts and managers with adaptive decision-making. We developed and implemented a Data 'fransformation System (DTS), a new Hybrid Genetic Algorithm (HGA) and an Anomaly Detection System (ADS) to address this challenge.
  • Thumbnail Image
    Item
    Open Access
    Link prediction and link detection in sequences of large social networks using temporal and local metrics
    (2006) Cooke, Richard Jeremy Edwin; Potgieter, Anet; April, Kurt
    This dissertation builds upon the ideas introduced by Liben-Nowell and Kleinberg in The Link Prediction Problem for Social Networks [42]. Link prediction is the problem of predicting between which unconnected nodes in a graph a link will form next, based on the current structure of the graph.
  • Thumbnail Image
    Item
    Open Access
    Towards a system redesign for better performance and customer satisfaction : a case study of the ICTS helpdesk at the University of Cape Town
    (2005) Balikuddembe, Joseph Kibombo; Potgieter, Anet
    This paper presents the findings from a study, which was carried out to investigate how the design of knowledge management systems could be improved for enhanced performance and greater customer satisfaction. The ICTS Department's helpdesk at the University of Cape Town, South Africa, was the venue for this case study. The study set out to meet the following objectives: - undertaking a knowledge acquisition strategy by carrying out a systems evaluation and analysis of the existing web-based user support system, - suggesting a knowledge representation model for an adaptive web-based user support system, and - developing and testing an online troubleshooter prototype for an improved knowledge use support system. To achieve the objectives of the study, knowledge engineering techniques were deployed on top of a qualitative research design. Questionnaires, which were supplemented by interview guides and observations, were the research tools used in gathering the data. In addition to this, a representative sample of the ICTS clientele and management was interviewed. It was discovered that poorly designed knowledge management systems cause frustration among the clientele who interact with the system. Specifically, it was found that the language used for knowledge representation plays a vital role in determining how best users can interpret knowledge items in a given knowledge domain. In other words, knowledge modelling and representation can improve knowledge representation if knowledge engineering techniques are appropriately followed in designing knowledge based systems. It was concluded that knowledge representation can be improved significantly if, firstly, the ontology technique is embraced as a mechanism of knowledge representation. Secondly, using hierarchies and taxonomies improves navigability in the knowledge structure. Thirdly, visual knowledge representation that supplements textual knowledge adds more meaning to the user, and is such a major and important technique that it can even cater for novice users.