Construction and optimisation of prediction intervals for neural network models

This thesis makes contributions to basic and fundamental research in the field of prediction interval construction using neural network models. It proposes novel methods for objective assessment, rapid construction, and optimisation of neural network-based prediction intervals for uncertainty quantification.

Bayesian model averaging of load demand forecasts from neural network models

Creating a set of a number of neural network (NN) models in an ensemble and accumulating them can achieve better overview capability as compared to single neural network. Neural network ensembles are designed to provide solutions to particular problems. Many researchers and academicians have adopted this NN ensemble technique, especially in machine learning, and has been applied in various fields of engineering, medicine and information technology. This paper present a robust aggregation methodology for load demand forecasting based on Bayesian Model Averaging of a set of neural network models in an ensemble. This paper estimate a vector of coefficient for individual NN models' forecasts using validation data-set. These coefficients, also known as weights, are equal to posterior probabilities of the models generating the forecasts. These BMA weights are then used in combining forecasts generated from NN models with test data-set. By comparing the Bayesian results with the Simple Averaging method, it was observed that benefits are obtained by utilizing an advanced method like BMA for forecast combinations.

A hybrid ART-GRNN online learning neural network with a ε-insensitive loss function

In this brief, a new neural network model called generalized adaptive resonance theory (GART) is introduced. GART is a hybrid model that comprises a modified Gaussian adaptive resonance theory (MGA) and the generalized regression neural network (GRNN). It is an enhanced version of the GRNN, which preserves the online learning properties of adaptive resonance theory (ART). A series of empirical studies to assess the effectiveness of GART in classification, regression, and time series prediction tasks is conducted. The results demonstrate that GART is able to produce good performances as compared with those of other methods, including the online sequential extreme learning machine (OSELM) and sequential learning radial basis function (RBF) neural network models.

Authenticating the identity of computer users with typing biometrics and the fuzzy min-max neural network

In this paper, typing biometrics is applied as an additional security measure to the password-based or Personal Identification Number (PIN)-based systems to authenticate the identity of computer users. In particular, keystroke pressure and latency signals are analyzed using the Fuzzy Min-Max (FMM) neural network for authentication purposes. A special pressure-sensitive keyboard is designed to collect keystroke pressure signals, in addition to the latency signals, from computer users when they type their passwords. Based on the keystroke pressure and latency signals, the FMM network is employed to classify the computer users into two categories, i.e., genuine users or impostors. To assess the effectiveness of the proposed approach, two sets of experiments are conducted, and the results are compared with those from statistical methods and neural network models. The experimental outcomes positively demonstrate the potentials of using typing biometrics and the FMM network to provide an additional security layer for the current password-based or PIN-based methods in authenticating the identity of computer users.

Uncertainty handling using neural network-based prediction intervals for electrical load forecasting

The complexity and level of uncertainty present in operation of power systems have significantly grown due to penetration of renewable resources. These complexities warrant the need for advanced methods for load forecasting and quantifying uncertainties associated with forecasts. The objective of this study is to develop a framework for probabilistic forecasting of electricity load demands. The proposed probabilistic framework allows the analyst to construct PIs (prediction intervals) for uncertainty quantification. A newly introduced method, called LUBE (lower upper bound estimation), is applied and extended to develop PIs using NN (neural network) models. The primary problem for construction of intervals is firstly formulated as a constrained single-objective problem. The sharpness of PIs is treated as the key objective and their calibration is considered as the constraint. PSO (particle swarm optimization) enhanced by the mutation operator is then used to optimally tune NN parameters subject to constraints set on the quality of PIs. Historical load datasets from Singapore, Ottawa (Canada) and Texas (USA) are used to examine performance of the proposed PSO-based LUBE method. According to obtained results, the proposed probabilistic forecasting method generates well-calibrated and informative PIs. Furthermore, comparative results demonstrate that the proposed PI construction method greatly outperforms three widely used benchmark methods. © 2014 Elsevier Ltd.

Improving an inverse model of sheet metal forming by neural network based regression

An inverse model for a sheet meta l forming process aims to determine the initial parameter levels required to form the final formed shape. This is a difficult problem that is usually approached by traditional methods such as finite element analysis. Formulating the problem as a classification problem makes it possible to use well established classification algorithms, such as decision trees. Classification is, however, generally based on a winner-takes-all approach when associating the output value with the corresponding class. On the other hand, when formulating the problem as a regression task, all the output values are combined to produce the corresponding class value. For a multi-class problem, this may result in very different associations compared with classification between the output of the model and the corresponding class. Such formulation makes it possible to use well known regression algorithms, such as neural networks. In this paper, we develop a neural network based inverse model of a sheet forming process, and compare its performance with that of a linear model. Both models are used in two modes, classification mode and a function estimation mode, to investigate the advantage of re-formulating the problem as a function estimation. This results in large improvements in the recognition rate of set-up parameters of a sheet metal forming process for both models, with a neural network model achieving much more accurate parameter recognition than a linear model.

An artificial neural network-based hairiness prediction model for worsted wool yarns

This study evaluated the performance of multilayer perceptron (MLP) and multivariate linear regression (MLR) models for predicting the hairiness of worsted-spun wool yarns from various top, yarn and processing parameters. The results indicated that the MLP model predicted yarn hairiness more accurately than the MLR model, and should have wide mill specific applications. On the basis of sensitivity analysis, the factors that affected yarn hairiness significantly included yarn twist, ring size, average fiber length (hauteur), fiber diameter and yarn count, with twist having the greatest impact on yarn hairiness.

Construction of neural network-based prediction intervals using particle swarm optimization

Prediction intervals (PIs) are excellent tools for quantification of uncertainties associated with point forecasts and predictions. This paper adopts and develops the lower upper bound estimation (LUBE) method for construction of PIs using neural network (NN) models. This method is fast and simple and does not require calculation of heavy matrices, as required by traditional methods. Besides, it makes no assumption about the data distribution. A new width-based index is proposed to quantitatively check how much PIs are informative. Using this measure and the coverage probability of PIs, a multi-objective optimization problem is formulated to train NN models in the LUBE method. The optimization problem is then transformed into a training problem through definition of a PI-based cost function. Particle swarm optimization (PSO) with the mutation operator is used to minimize the cost function. Experiments with synthetic and real-world case studies indicate that the proposed PSO-based LUBE method can construct higher quality PIs in a simpler and faster manner.

A new neural network-based type reduction algorithm for interval type-2 fuzzy logic systems

This paper introduces a new type reduction (TR) algorithm for interval type-2 fuzzy logic systems (IT2 FLSs). Flexibility and adaptiveness are the key features of the proposed non-parametric algorithm. Lower and upper firing strengths of rules as well as their consequent coefficients are fed into a neural network (NN). NN output is a crisp value that corresponds to the defuzzified output of IT2 FLSs. The NN type reducer is trained through minimization of an error-based cost function with the purpose of improving modelling and forecasting performance of IT2 FLS models. Simulation results indicate that application of the proposed TR algorithm greatly enhances modelling and forecasting performance of IT2 FLS models. This benefit is achieved in no cost, as the computational requirement of the proposed algorithm is less than or at most equivalent to traditional TR algorithms.

Neural network and interval type-2 fuzzy system for stock price forecasting

Stock price forecast has long been received special attention of investors and financial institutions. As stock prices are changeable over time and increasingly uncertain in modern financial markets, their forecasting becomes more important than ever before. A hybrid approach consisting of two components, a neural network and a fuzzy logic system, is proposed in this paper for stock price prediction. The first component of the hybrid, i.e. a feedforward neural network (FFNN), is used to select inputs that are highly relevant to the dependent variables. An interval type-2 fuzzy logic system (IT2 FLS) is employed as the second component of the hybrid forecasting method. The IT2 FLS’s parameters are initialized through deployment of the k-means clustering method and they are adjusted by the genetic algorithm. Experimental results demonstrate the efficiency of the FFNN input selection approach as it reduces the complexity and increase the accuracy of the forecasting models. In addition, IT2 FLS outperforms the widely used type-1 FLS and FFNN models in stock price forecasting. The combination of the FFNN and the IT2 FLS produces dominant forecasting accuracy compared to employing only the IT2 FLSs without the FFNN input selection.

Neural network ensemble : evaluation of aggregation algorithms for electricity demand forecasting

This paper examines and analyzes different aggregation algorithms to improve accuracy of forecasts obtained using neural network (NN) ensembles. These algorithms include equal-weights combination of Best NN models, combination of trimmed forecasts, and Bayesian Model Averaging (BMA). The predictive performance of these algorithms are evaluated using Australian electricity demand data. The output of the aggregation algorithms of NN ensembles are compared with a Naive approach. Mean absolute percentage error is applied as the performance index for assessing the quality of aggregated forecasts. Through comprehensive simulations, it is found that the aggregation algorithms can significantly improve the forecasting accuracies. The BMA algorithm also demonstrates the best performance amongst aggregation algorithms investigated in this study.

Variance-covariance based weighing for neural network ensembles

Neural network (NN) is a popular artificial intelligence technique for solving complicated problems due to their inherent capabilities. However generalization in NN can be harmed by a number of factors including parameter's initialization, inappropriate network topology and setting parameters of the training process itself. Forecast combinations of NN models have the potential for improved generalization and lower training time. A weighted averaging based on Variance-Covariance method that assigns greater weight to the forecasts producing lower error, instead of equal weights is practiced in this paper. While implementing the method, combination of forecasts is done with all candidate models in one experiment and with the best selected models in another experiment. It is observed during the empirical analysis that forecasting accuracy is improved by combining the best individual NN models. Another finding of this study is that reducing the number of NN models increases the diversity and, hence, accuracy.

Artificial neural network analysis of twin tunnelling-induced ground settlements

In this paper, we apply a computational intelligence method for tunnelling settlement prediction. A supervised feed forward back propagation neural network is used to predict the surface settlement during twin-tunnelling while surface buildings are considered in the models. The performance of the statistical neural network structure is tested on a dataset provided by numerical parametric studies conducted by ABAQUS software based on Shiraz line 1 metro data. Six input variables are fed to neural network model for predicting the surface settlement. These include tunnel center depth, distance between centerlines of twin tunnels, buildings width and building bending stiffness, and building weight and distance to tunnel centerline. Simulation results indicate that the proposed NN models are able to accurately predict the surface settlement.