List of Titles

Quick View

A computational intelligence based prediction model for flight departure delays

- Hopane, Johanna Mmakwena

Authors: Hopane, Johanna Mmakwena
Date: 2019
Subjects: Computational intelligence , Flights delays - South Africa , OR Tambo International Airport
Language: English
Type: Masters (Thesis)
Identifier: http://hdl.handle.net/10210/403143 , uj:33768
Description: Abstract : Flight departure delays are a major problem at OR Tambo International airport (ORTIA). There is a high delay for flights to depart, especially at the beginning of the month and at the end of the month. The increasing demand for flights departing at ORTIA often leads to a negative effect on business deals, individuals’ health, job opportunities and tourists. When flights are delayed departing, travellers are notified at the airport every 30 minutes about the status of the flight and the reason the flight is delayed if it is known. This study aims to construct a flight delays prediction model using machine learning algorithms. The flight departures data were obtained from ORTIAs website timetable for departing flight schedules. The flight departure data for ORTIA to any destination (i.e. Johannesburg (JNB) Airport to Cape Town (CPT)) for South African Airways (SAA) airline was used for this study. Machine learning algorithms namely Decision Trees (J48), Support Vector Machine (SVM), K-Means Clustering (K-Means) and Multi-Layered Perceptron (MLP) were used to construct the flight departure delays prediction models. A cross-validation (CV) method was used for evaluating the models. The best prediction model was selected by using a confusion matrix. The results showed that the models constructed using Decision Trees (J48) achieved the best prediction for flight departure delays at 67.144%, while Multi-layered Perceptron (MLP) obtained 67.010%, Support Vector Machine (SVM) obtained 66.249% and K-Means Clustering (K-Means) obtained 61.549%. Travellers wishing to travel from ORTIA can predict flight departure delays using this tool. This tool will allow travellers to enter variables such as month, week of month, day of week and time of day. The entered variables will predict the flight departure status by examining target concepts such as On Time, Delayed and Cancelled. The travellers will only be able to predict flight departures status, although they will not have full knowledge of the flight departures volume. In that case, they will depend on the flight information display system (FIDS) board. This study can predict and empower travellers by providing them with a tool that can determine the punctuality of the flights departing from ORTIA. , M.Com. (Information Technology Management)
Full Text:

Quick View

A distributed, multi-agent model for general purpose crowd simulation

- Ekron, Kieron Charles

Authors: Ekron, Kieron Charles
Date: 2012-11-06
Subjects: Multiagent systems , Crowd simulation , Computational intelligence , Intelligent agents (Computer software)
Type: Thesis
Identifier: uj:7362 , http://hdl.handle.net/10210/8119
Description: M.Sc. (Computer Science) , The purpose of the research presented in this dissertation is to explore the use of a distributed multi-agent system in a general purpose crowd simulation model. Crowd simulation is becoming an increasingly important tool for analysing new construction projects, as it enables safety and performance evaluations to be performed on architectural plans before the buildings have been constructed. Crowd simulation is a challenging problem, as it requires the simulation of complex interactions of people within a crowd. The dissertation investigates existing models of crowd simulation and identifies three primary sub-tasks of crowd simulation: deliberation, path planning and collision-avoiding movement. Deliberation is the process of determining which goal an agent will attempt to satisfy next. Path planning is the process of finding a collision-free path from an agent‟s current location towards its goal. Collision-avoiding movement deals with moving an agent along its calculated path while avoiding collisions with other agents. A multi-agent crowd simulation model, DiMACS, is proposed as a means of addressing the problem of crowd simulation. Multi-agent technology provides an effective solution for representing individuals within a crowd; each member of a crowd can be represented as an intelligent agent. Intelligent agents are capable of maintaining their own internal state and deciding on a course of action based on that internal state. DiMACS is capable of producing realistic simulations while making use of distributed and parallel processing to improve its performance. In addition, the model is highly customisable. The dissertation also presents a user-friendly method for configuring agents within a simulation that abstracts the complexity of agent behaviour away from a user so as to increase the accessibility of configuring the proposed model. In addition, an application programming interface is provided that enables developers to extend the model to simulate additional agent behaviours. The research shows how distributed and parallel processing may be used to improve the performance of an agent-based crowd simulation without compromising the accuracy of the simulation.
Full Text:

Quick View

A rough set approach to bushings fault detection

- Mpanza, Lindokuhle Justice

Authors: Mpanza, Lindokuhle Justice
Date: 2012-06-06
Subjects: Bushings , Electric insulators , Fault detection methods , Rough set theory , Ant colony optimization , Computational intelligence , Dissolved gas analysis
Type: Thesis
Identifier: uj:2502 , http://hdl.handle.net/10210/4955
Description: M. Ing. , Fault detection tools have gained popularity in recent years due to the increasing need for reliable and predictable equipments. Transformer bushings account for the majority of transformer faults. Hence, to uphold the integrity of the power transmission and dis- tribution system, a tool to detect and identify faults in their developing stage is necessary in transformer bushings. Among the numerous tools for bushings monitoring, dissolved gas analysis (DGA) is the most commonly used. The advances in DGA and data storage capabilities have resulted in large amount of data and ultimately, the data analysis crisis. Consequent to that, computational intelligence methods have advanced to deal with this data analysis problem and help in the decision-making process. Numerous computational intelligence approaches have been proposed for bushing fault detection. Most of these approaches focus on the accuracy of prediction and not much research has been allocated to investigate the interpretability of the decisions derived from these systems. This work proposes a rough set theory (RST) model for bushing fault detection based on DGA data analyzed using the IEEEc57.104 and the IEC 60599 standards. RST is a rule-based technique suitable for analyzing vague, uncertain and imprecise data. RST extracts rules from the data to model the system. These rules are used for prediction and interpreting the decision process. The lesser the number of rules, the easier it is to interpret the model. The performance of the RST is dependent on the discretization technique employed. An equal frequency bin (EFB), Boolean reasoning (BR) and entropy partition (EP) are used to develop an RST model. The model trained using EFB data performs better than the models trained using BR and EP. The accuracy achieved is 96.4%, 96.0% and 91.3% for EFB, BR and EP respectively. This work also pro poses an ant colony optimization (ACO) for discretization. A model created using ACO discretized achieved an accuracy of 96.1%, which is compatible with the three methods above. When considering the overall performance, the ACO is a better discretization tool since it produces an accurate model with the least number of rules. The rough set tool proposed in this work is benchmarked against a multi-layer perceptron (MLP) and radial basis function (RBF) neural networks. Results prove that RST modeling for bushing is equally as capable as the MLP and better than RBF. The RST, MLP and RBF are used in an ensemble of classifiers. The ensemble performs better than the standalone models.
Full Text:

Quick View

Computational intelligence technology for the generation of building layouts combined with multi-agent furniture placement

- Bijker, Jacobus Jan

Authors: Bijker, Jacobus Jan
Date: 2012-11-02
Subjects: Multiagent systems , Computational intelligence , Computer-aided engineering , Architectural design
Type: Thesis
Identifier: uj:7318 , http://hdl.handle.net/10210/8056
Description: M.Sc. (Computer Science) , This dissertation presents a method for learning from existing building designs and generating new building layouts. Generating fully furnished building layouts could be very useful for video games or for assisting architects when designing new buildings. The core concern is to drastically reduce the workload required to design building layouts. The implemented prototype features a Computer Aided Design system, named CABuilD that allows users to design fully furnished multi-storey building layouts. Building layouts designed using CABuilD can be taught to an Artificial Immune System. The Artificial Immune System tracks information such as building layouts, room sizes and furniture layouts. Once building layouts has been taught to the artificial immune system, a generation algorithm can utilise the information in order to generate fully furnished building layouts. The generation algorithm that is presented allows fully furnished buildings to be generated from high-level information such as the number of rooms to include and a building perimeter. The presented algorithm differs from existing building generation methods in the following ways: Firstly existing methods either ignore building perimeters or assume a buildings perimeter is a rectangle. The presented method allows the user to specify a closed polygon as a building perimeter which will guide the generation of the building layout. Secondly existing generation methods tend to run from a set of rules. The implemented system learns from existing building layouts, effectively allowing it to generate different building types based on the building layouts that were taught to the system. Thirdly, the system generates both the building layout as well as the furniture within rooms. Existing systems only generate the building layout or the furniture, but not both. The prototype that was implemented as a proof of concept uses a number of biologically inspired techniques such as Ant algorithms, Particle Swarm Optimisation and Artificial Immune Systems. The system also employs multiple intelligent agents in order to furnished rooms. The prototype is capable of generating furnished building layouts in merely a few seconds, much faster than a human could design such a layout. Possible improvements and future work is presented at the end of the dissertation.
Full Text:

Quick View

Condition monitoring of transformer’s bushings using computational intelligence

- Maumela, Joshua Tshifhiwa

Authors: Maumela, Joshua Tshifhiwa
Date: 2014-04-16
Subjects: Electric insulators and insulation , Electric power systems , Intelligent agents (Computer software) , Computational intelligence
Type: Thesis
Identifier: uj:10800 , http://hdl.handle.net/10210/10307
Description: M.Ing. (Electrical and Electronic Engineering) , Dissolved Gas-in-oil analysis (DGA) is used to monitor the condition of bushings on large power transformers. There are different techniques used in determining the conditions from the data collected, but in this work the Artificial Intelligence techniques are investigated. This work investigates which gases in DGA are related to each other and which ones are important for making decisions. When the related and crucial gases are determined, the other gases are discarded thereby reducing the number of attributes in DGA. Hence a further investigation is done to see how these new datasets influence the performance of the classifiers used to classify the DGA of full attributes. The classifiers used in these experiments were Backpropagation Neural Networks (BPNN) and Support Vector Machines (SVM) whereas the Principal Component Analysis (PCA), Rough Set (RS), Incremental Granular Ranking (GR++) and Decision Trees (DT) were used to reduce the attributes of the dataset. The parameters used when training the BPNN and SVM classifiers are kept fixed to create a controlled test environment when investigating the effects of reducing the number of gases. This work further introduced a new classifier that can handle high dimension dataset and noisy dataset, Rough Neural Network (RNN). This classifier was tested when trained using the full dataset and how it is affected by reducing the number of gases used to train it. The results in these experiments showed that ethane and total combustible gases attributes are core attributes chosen by the four algorithms as gases needed for decision making. The average results of the classification performance showed that the reduction of attributes helps improve the performance of classifiers. Hence the science of transformer condition monitoring can be derived from studying the relations and patterns created by the different gases attributes in DGA. This statement is supported by the classification improvements where the RNN classifier had 99.7% classification accuracy when trained using the three attributes determined by the PCA.
Full Text:

Quick View

Development of effective cuckoo search algorithms for optimisation purposes

- Mareli, Mahlaku

Authors: Mareli, Mahlaku
Date: 2018
Subjects: Computational intelligence , Computer algorithms , Mathematical optimization - Data processing , Computer networks
Language: English
Type: Doctoral (Thesis)
Identifier: http://hdl.handle.net/10210/295009 , uj:32110
Description: Abstract: Optimisation, the process of finding either a maximum of a minimum of the problem at hand plays a key role in several disciplines including engineering and science. In this thesis, different Cuckoo Search algorithms are developed for effective optimisation purposes. These algorithms are tested on ten mathematical test functions and then used to optimise a Back-Propagation Neural Network used for short-term electricity load forecasting for South African data, with the focus on the City of Johannesburg. The original Cuckoo Search algorithm is based on random walk step sizes derived from Lévy probability distribution and the switching parameter between local and global random walks is constant. However, other probability distributions like Cauchy, Gaussian and Gamma have also been used and the switching parameter can be changed dynamically. The first contribution of the thesis is the development a new Cuckoo Search algorithm whose random step sizes are derived from Pareto probability distribution function. This new Pareto-based Cuckoo Search algorithm is tested on ten benchmark test functions together with other Cuckoo Search algorithms using step sizes derived from Gaussian, Cauchy, Gamma and Lévy probability density functions. When using the confidence interval analysis, the Lévy-based Cuckoo Search algorithm outperforms the Pareto based Cuckoo. However, confidence interval results are only superior due to only one test function whereby Lévy-based Cuckoo Search performed well. Moreover, the Pareto-based Cuckoo shows superior performance in comparison to the other algorithms, leading in seven test functions out of ten when tested for convergence. The second contribution is the implementation of Cuckoo Search algorithms with dynamically increasing switching parameters between local and random walks. The first improvement done on Cuckoo Search algorithm is the implementation of linear increasing switching parameter, the second is the implementation of power increasing switching parameter and the third improvement is the implementation of exponential increasing switching parameter. When tested on benchmark test functions, the exponentially increasing Cuckoo Search algorithm outperforms the other algorithms by obtaining the longest confidence interval of 4.50566 while the next algorithm (original Cuckoo Search) obtains an interval of 3.9699. Moreover, using convergence plots, both... , D.Ing. (Electrical and Electronic Engineering)
Full Text:

Quick View

Economic modelling using computational intelligence techniques

- Khoza, Msizi Smiso

Authors: Khoza, Msizi Smiso
Date: 2013-12-09
Subjects: Computational intelligence , Economic modelling , Economics, Mathematical , Economics - Data processing , Economics - Computer simulation , Knowledge acquisition (Expert systems)
Type: Thesis
Identifier: uj:7809 , http://hdl.handle.net/10210/8704
Description: M.Ing. ( Electrical & Electronic Engineering Science) , Economic modelling tools have gained popularity in recent years due to the increasing need for greater knowledge to assist policy makers and economists. A number of computational intelligence approaches have been proposed for economic modelling. Most of these approaches focus on the accuracy of prediction and not much research has been allocated to investigate the interpretability of the decisions derived from these systems. This work proposes the use of computational intelligence techniques (Rough set theory (RST) and the Multi-layer perceptron (MLP) model) to model the South African economy. RST is a rule-based technique suitable for analysing vague, uncertain and imprecise data. RST extracts rules from the data to model the system. These rules are used for prediction and interpreting the decision process. The lesser the number of rules, the easier it is to interpret the model. The performance of the RST is dependent on the discretization technique employed. An equal frequency bin (EFB), Boolean reasoning (BR), entropy partition (EP) and the Naïve algorithm (NA) are used to develop an RST model. The model trained using EFB data performs better than the models trained using BR and EP. RST was used to model South Africa’s financial sector. Here, accuracy of 86.8%, 57.7%, 64.5% and 43% were achieved for EFB, BR, EP and NA respectively. This work also proposes an ensemble of rough set theory and the multi-layer perceptron model to model the South African economy wherein, a prediction of the direction of the gross domestic product is presented. This work also proposes the use of an auto-associative Neural Network to impute missing economic data. The auto-associative neural network imputed the ten variables or attributes that were used in the prediction model. These variables were: Construction contractors rating lack of skilled labour as constraint, Tertiary economic sector contribution to GDP, Income velocity of circulation of money, Total manufacturing production volume, Manufacturing firms rating lack of skilled labour as constraint, Total asset value of banking industry, Nominal unit labour cost, Total mass of Platinum Group Metals (PGMs) mined, Total revenue from sale of PGMs and the Gross Domestic Expenditure (GDE). The level of imputation accuracy achieved varied with the attribute. The accuracy ranged from 85.9% to 98.7%.
Full Text:

Quick View

Empirical evaluation of optimization techniques for classification and prediction tasks

- Leke, Collins Achepsah

Authors: Leke, Collins Achepsah
Date: 2014-03-27
Subjects: Genetic algorithms , Statistical decision - Data processing , Decision support systems , Missing observations (Statistics) , Computational intelligence
Type: Thesis
Identifier: uj:4520 , http://hdl.handle.net/10210/9858
Description: M.Ing. (Electrical and Electronic Engineering) , Missing data is an issue which leads to a variety of problems in the analysis and processing of data in datasets in almost every aspect of day−to−day life. Due to this reason missing data and ways of handling this problem have been an area of research in a variety of disciplines in recent times. This thesis presents a method which is aimed at finding approximations to missing values in a dataset by making use of Genetic Algorithm (GA), Simulated Annealing (SA), Particle Swarm Optimization (PSO), Random Forest (RF), Negative Selection (NS) in combination with auto-associative neural networks, and also provides a comparative analysis of these algorithms. The methods suggested use the optimization algorithms to minimize an error function derived from training an auto-associative neural network during which the interrelationships between the inputs and the outputs are obtained and stored in the weights connecting the different layers of the network. The error function is expressed as the square of the difference between the actual observations and predicted values from an auto-associative neural network. In the event of missing data, all the values of the actual observations are not known hence, the error function is decomposed to depend on the known and unknown variable values. Multi Layer Perceptron (MLP) neural network is employed to train the neural networks using the Scaled Conjugate Gradient (SCG) method. The research primarily focusses on predicting missing data entries from two datasets being the Manufacturing dataset and the Forest Fire dataset. Prediction is a representation of how things will occur in the future based on past occurrences and experiences. The research also focuses on investigating the use of this proposed technique in approximating and classifying missing data with great accuracy from five classification datasets being the Australian Credit, German Credit, Japanese Credit, Heart Disease and Car Evaluation datasets. It also investigates the impact of using different neural network architectures in training the neural network and finding approximations for the missing values, and using the best possible architecture for evaluation purposes. It is revealed in this research that the approximated values for the missing data obtained by applying the proposed models are accurate with a high percentage of correlation between the actual missing values and corresponding approximated values using the proposed models on the Manufacturing dataset ranging between 94.7% and 95.2% with the exception of the Negative Selection algorithm which resulted in a 49.6% correlation coefficient value. On the Forest Fire dataset, it was observed that there was a low percentage correlation between the actual missing values and the corresponding approximated values in the range 0.95% to 4.49% due to the nature of the values of the variables in the dataset. The Negative Selection algorithm on this dataset revealed a negative percentage correlation between the actual values and the approximated values with a value of 100%. Approximations found for missing data are also observed to depend on the particular neural network architecture employed in training the dataset. Further analysis revealed that the Random Forest algorithm on average performed better than the GA, SA, PSO, and NS algorithms yielding the lowest Mean Square Error, Root Mean Square Error, and Mean Absolute Error values. On the other end of the scale was the NS algorithm which produced the highest values for the three error metrics bearing in mind that for these, the lower the values, the better the performance, and vice versa. The evaluation of the algorithms on the classification datasets revealed that the most accurate in classifying and identifying to which of a set of categories a new observation belonged on the basis of the training set of data is the Random Forest algorithm, which yielded the highest AUC percentage values on all of the five classification datasets. The differences between its AUC values and those of the GA, SA, PSO, and NS algorithms were statistically significant, with the most statistically significant differences observed when the AUC values for the Random Forest algorithm were compared to those of the Negative Selection algorithm on all five classification datasets. The GA, SA, and PSO algorithms produced AUC values which when compared against each other on all five classification datasets were not very different. Overall analysis on the datasets considered revealed that the algorithm which performed best in solving both the prediction and classification problems was the Random Forest algorithm as seen by the results obtained. The algorithm on the other end of the scale after comparisons of results was the Negative Selection algorithm which produced the highest error metric values for the prediction problems and the lowest AUC values for the classification problems.
Full Text:

Quick View

Intelligent pre-processing for data mining

- De Bruin, Ludwig

Authors: De Bruin, Ludwig
Date: 2014-06-26
Subjects: Data mining , Intelligent agents (Computer software) , Computational intelligence
Type: Thesis
Identifier: uj:11611 , http://hdl.handle.net/10210/11324
Description: M.Sc. (Information Technology) , Data is generated at an ever-increasing rate and it has become difficult to process or analyse it in its raw form. The most data is generated by processes or measuring equipment, resulting in very large volumes of data per time unit. Companies and corporations rely on their Management and Information Systems (MIS) teams to perform Extract, Transform and Load (ETL) operations to data warehouses on a daily basis in order to provide them with reports. Data mining is a Business Intelligence (BI) tool and can be defined as the process of discovering hidden information from existing data repositories. The successful operation of data mining algorithms requires data to be pre-processed for algorithms to derive IF-THEN rules. This dissertation presents a data pre-processing model to transform data in an intelligent manner to enhance its suitability for data mining operations. The Extract Pre- Process and Save for Data Mining (EPS4DM) model is proposed. This model will perform the pre-processing tasks required on a chosen dataset and transform the dataset into the formats required. This can be accessed by data mining algorithms from a data mining mart when needed. The proof of concept prototype features agent-based Computational Intelligence (CI) based algorithms, which allow the pre-processing tasks of classification and clustering as means of dimensionality reduction to be performed. The task of clustering requires the denormalisation of relational structures and is automated using a feature vector approach. A Particle Swarm Optimisation (PSO) algorithm is run on the patterns to find cluster centres based on Euclidean distances. The task of classification requires a feature vector as input and makes use of a Genetic Algorithm (GA) to produce a transformation matrix to reduce the number of significant features in the dataset. The results of both the classification and clustering processes are stored in the data mart.
Full Text:

Quick View

Optimization of PID controller with metaheuristic algorithms for DC motor drives : review

- Oladipo, Stephen, Sun, Yanxia, Wang, Zenghui

Authors: Oladipo, Stephen , Sun, Yanxia , Wang, Zenghui
Date: 2020
Subjects: oportional-Integral-Derivate , Metaheuristic algorithm , Computational intelligence
Language: English
Type: Article
Identifier: http://hdl.handle.net/10210/455182 , uj:40278 , Citation: Oladipo, S., Sun, Y. & Wang, Z. 2020. Optimization of PID controller with metaheuristic algorithms for DC motor drives : review.
Description: Abstract: DC motor is extensively used in various industrial applications such as robotics, automobiles, toys and many other motoring applications. This is attributable to their extraordinary flexibility, durability and low implementation cost. To obtain the desired output based on the use of the DC motor, it is imperative to control the speed, position, torque and other variables of the DC motor. Many classical techniques have been utilized in the past to control the DC motor, however, such classical methods typically take a long time, particularly when used for complex nonlinear systems. The use of metaheuristic algorithms as a way of implementing artificial intelligence (AI) in this field has proven to be highly effective in overcoming these shortcomings. In recent decades, metaheuristic algorithms have become increasingly prevalent due to their tremendous success in addressing several real-world optimization challenges in various areas of human activities, ranging from economic, pharmaceutical and industrial applications to intellectual applications. This review presents the use of different types of metaheuristic algorithm techniques in optimizing the parameters of the proportional-integralderived (PID) controller in order to control the DC motor. For a more robust review, the application of various forms of PID controller, as well as different types of DC motors, is considered.
Full Text:

Quick View

Showing items 1 - 14 of 14