In recent years, several new methods for missing data estimation have been developed. Real world datasets possess the properties of big data being volume, velocity and variety. With an increase in volume which includes sample size and dimensionality, existing imputation methods have become less effective and accurate. Much attention has been given to narrow Artificial Intelligence frameworks courtesy of their efficiency in low dimensional settings. However, with an increase in dimensionality, these methods yield unrepresentative imputations with an impact on decision making processes. Therefore in this paper, we present a new framework for missing data imputation in high dimensional datasets. A Deep Learning technique is used in conjunction with a swarm intelligence algorithm. The performance of the proposed technique was experimentally tested and compared against other existing methods on an off-line dataset. The results obtained have shown promising potential with slightly longer execution times, which are a worthy tradeoff when accuracy is of importance.

From inception, side channel leakage has been widely used for the purposes of extracting secret information, such as cryptographic keys, from embedded devices. However, in a few instances it has been utilized for extracting other information about the internal state of a computing device. In this paper, we exploit side channel information to recover large parts of the Sykipot malware program executed on a smart card. We present the first methodology to recover the program code of a smart card malware by evaluating its power consumption only. Besides well-studied methods from side channel analysis, we apply a combination of dimensionality reduction techniques in the form of PCA and LDA models to compress the large amount of data generated while preserving as much variance of the original data as possible. Among feature extraction techniques, PCA and LDA are very common dimensionality reduction algorithms that have successfully been applied in many classification problems like face recognition, character recognition, speech recognition, etc. with the chief objective being to eliminate insignificant data (without losing too much information) during the pre-processing step. In addition to quantifying the potential of the created side channel based disassembler, we highlight its diverse and unique application scenarios.

In this paper it presents a methodology that aims to deliver a near optimal Distributed Generation (DG) in the process of allocation DG units by using a hybrid genetic algorithm, Hence, the main aim is to minimize power losses in DG. The proposed algorithm in this paper involves two main parts of algorithms an artificial neural network (ANN) that found to evaluate the fitness function in the generation of the best DG distribution and a local search procedure that allows the algorithm to search a massive range of neighbours units.

In this paper a comparison between an ensembles (multi-classifier) constructed of several machine learning methods (support vector machine, artificial neural network, naive Bayesian classifier, decision trees, radial basis function and k nearest neighbors) versus each single classifiers of these methods in term of gold mine underground dam levels prediction is presented. The ensembles as well as the single classifiers are used to classify, thus monitoring and predicting the underground water dam levels on a single-pump station deep gold in South Africa. In order to improve the classification accuracy an ensemble was constructed based on each single classifier performance, therefore, five ensembles were built and tested. In terms of misclassification error, the results show the ensemble to be more efficient for classification of underground water dam levels compared to each of the single classifiers.

This paper presents an experiment that consists of constructing auto-regressive moving average (ARMA), neural networks and neuro-fuzzy models with historical electricity consumption time series data to create models that can be used to forecast consumption in the future. The data was sampled on a monthly basis from January 1985 to December 2011. An ARMA, multilayer perceptron neural network with back propagation and neuro-fuzzy modelling technique which combines Takagi-Sugeno fuzzy models and neural networks were used to create the models for one step ahead forecasting. The results of the three techniques were compared and the results show that neuro-fuzzy models outperformed the neural network and ARMA models in terms of accuracy.

Four optimization algorithms (genetic algorithm, simulated annealing, particle swarm optimization and random forest) were applied with an MLP based auto associative neural network on two classification datasets and one prediction dataset. This work was undertaken to investigate the effectiveness of using auto associative neural networks and optimization algorithms in missing data prediction and classification tasks. If performed appropriately, computational intelligence and optimization algorithm systems could lead to consistent, accurate and trustworthy predictions and classifications resulting in more adequate decisions. The results reveal GA, SA and PSO to be more efficient when compared to RF in terms of predicting the forest area to be affected by fire. GA, SA, and PSO had the same accuracy of 93.3%, while RF showed 92.99% accuracy. For the classification problems, RF showed 93.66% and 92.11% accuracy on the German credit and Heart disease datasets respectively, outperforming GA, SA and PSO.

Automobile industry is an integral component of the South African economy. The industry contributes 6 to 7% to the country's economy and is occupied by major international automobile companies. In this highly competitive industry, insights into overall after sales customer satisfaction is of great value to the companies to better serve customers and retain their competitive advantage. Following established researches in the field of customer satisfaction, this paper views overall after sales customer satisfaction as a function of the various dimensions of service quality. Subsequently Multinomial Logistic Regression is used to study the effect of these dimensions of service quality on the overall after sales customer satisfaction. The results show that customers value more the technical and behavioral competencies of the service staff, rather than the physical infrastructure and physical appearance of the service staffs.

Abstract Motivation Road traffic accidents are among the top leading causes of deaths and injuries of various levels in South Africa. With the wealth and huge amount of data generated from road traffic accidents, the issue of traffic accident prediction has become a central challenge in the field of transportation data analysis. Such accident prediction is designed to detect patterns involved in dangerous crashes and thus help decision making and planning before casualty and loss occur. Recently, numerous researchers have presented a wide range of prediction techniques. Most of these methods are based on statistical studies but usually fail to explain the insights of prediction results. This has led to the development and application of supervised learning algorithms (classifiers) in an attempt to provide more accurate accident prediction in terms of injury severity (fatal/serious/slight/property damage with no injury). Even then, the task of learning an accurate classifier from instances raises a number of new issues some of which have not been properly addressed by transportation research. Thus, an effective prediction method is required for improving predictive accuracy. Results The essence of the paper is the proposal that prediction of accidents given poor data quality (in terms of incomplete data) can be improved by using a classifier based on grey relational analysis, a similarity‐based method. We evaluate the grey relational classifier with other state‐of‐the‐art classifiers including artificial neural networks, classification and regression trees, k‐nearest neighbour, linear discriminant analysis, naïve Bayes classifier, algorithm quasi‐optimal and support vector machines. Real‐world road traffic accident dataset is utilized for this task. Experimental results are provided to illustrate the efficiency and the robustness of the grey relational classifier algorithm in terms of road traffic accident predictive accuracy.

In this study we propose a multilayered feedforward neural network (MFNN) with Bayesian Regularization, and apply it to the credit risk evaluation problem domain using a real world data set from a financial services company in England. We choose the MFNN because of its broad applicability to many problem domains of relevance to business: principally prediction, classification, and modelling. We employ two different methods to determine their prowess in identifying the true positives, that is, defaulters. We analyzed the effect of making the number of observed bad equal the number of observed good in the data by over sampling of the minority class (bad obligors) by resampling without replacement, and compare this to the dimensionality reduction of the input vector space using Principal Component Analysis. Overall results indicate that using the Receiver Operating Characteristic as a measure of discriminatory power, over sampling of the minority class has been found to be effective in identifying the true positives.

This paper presents an experiment that consists of constructing neural networks and neuro-fuzzy models with historical electricity consumption time series data to create models that can be used to forecast consumption in the future. The data was sampled on a monthly basis from January 1985 to December 2011. A multilayer perceptron neural network with back propagation and neuro-fuzzy modelling technique which combines Takagi-Sugeno fuzzy models and neural networks were used to create the models. Different input sizes were used to construct 10 different one step ahead forecasting models for each technique. The results of the two techniques were compared and the results show that neuro-fuzzy models outperformed the neural network models in terms of accuracy.

Machine learning has been successfully used for credit-evaluation decisions. Most research on machine learning assumes that the attributes of training and tests instances are not only completely specified but are also free from noise. Real world data

The nature of missing data problems forces us to build models that maintain high accuracies and steadiness. The models developed to achieve this are usually complex and computationally expensive. In this paper, we propose an unsupervised multi-layered artificial immune system for an insurance classification problem that is characterised as highly dimensional and contains escalating missing data. The system is compared with the k-nearest neighbour, support vector machines and logistic discriminant models. Overall, the results show that whilst k-nearest neighbour achieves the highest accuracy, the multi-layered artificial immune system is steady and maintains high performance compared to other models, regardless of how the missing data is distributed in a dataset.

Constructing an accurate effort prediction model is a challenge in software engineering. The development and validation of models that are used for prediction tasks require good quality data. Unfortunately, software engineering datasets tend to suffer from the incompleteness which could result to inaccurate decision making and project management and implementation. Recently, the use of machine learning algorithms has proven to be of great practical value in solving a variety of software engineering problems including software prediction, including the use of ensemble (combining) classifiers. Research indicates that ensemble individual classifiers lead to a significant improvement in classification performance by having them vote for the most popular class. This paper proposes a method for improving software effort prediction accuracy produced by a decision tree learning algorithm and by generating the ensemble using two imputation methods as elements. Benchmarking results on ten industrial datasets show that the proposed ensemble strategy has the potential to improve prediction accuracy compared to an individual imputation method, especially if multiple imputation is a component of the ensemble.

Data quality is an important aspect of empirical analysis. This paper compares three noise handling methods to assess the benefit of identifying and either filtering or editing problematic instances. We compare a 'do nothing' strategy with (i) filtering, (ii) robust filtering and (Hi) filtering followed by polishing. A problem is that it is not possible to determine whether an instance contains noise unless it has implausible values. Since we cannot determine the true overall noise level we use implausible val.ues as a proxy measure. In addition to the ability to identify implausible values, we use another proxy measure, the ability to fit a classification tree to the data. The interpretation is low misclassification rates imply low noise levels. We found that all three of our data quality techniques improve upon the 'do nothing' strategy, also that the filtering and polishing was the most effective technique for dealing with noise since we eliminated the fewest data and had the lowest misclassification rates. Unfortunately the polishing process introduces new implausible values. We believe consideration of data quality is an important aspect of empirical software engineering. We have shown that for one large and complex real world data set automated techniques can help isolate noisy instances and potentially polish the values to produce better quality data for the analyst. However this work is at a preliminary stage and it assumes that the proxy measures of lity are appropriate.

OBJECTIVE – The aim is to report upon an assessment of the impact noise has on the predictive accuracy by comparing noise handling techniques. METHOD – We describe the process of cleaning a large software management dataset comprising initially of more than 10,000 projects. The data quality is mainly assessed through feedback from the data provider and manual inspection of the data. Three methods of noise correction (polishing, noise elimination and robust algorithms) are compared with each other assessing their accuracy. The noise detection was undertaken by using a regression tree model. RESULTS – Three noise correction methods are compared and different results in their accuracy where noted. CONCLUSIONS – The results demonstrated that polishing improves classification accuracy compared to noise elimination and robust algorithms approaches.

Software engineers are commonly faced with the problem of incomplete data. Incomplete data can reduce system performance in terms of predictive accuracy. Unfortunately, rare research has been conducted to systematically explore the impact of missing values, especially from the missing data handling point of view. This has made various missing data techniques (MDTs) less significant. This paper describes a systematic comparison of seven MDTs using eight industrial datasets. Our findings from an empirical evaluation suggest listwise deletion as the least effective technique for handling incomplete data while multiple imputation achieves the highest accuracy rates. We further propose and show how a combination of MDTs by randomizing a decision tree building algorithm leads to a significant improvement in prediction performance for missing values up to 50%.