No abstract

Artificial neural networks provide a promising alternative to hydrological time series modelling. However, there are still many fundamental problems requiring further analyses, such as structure identification, parameter estimation, generalization, performance improvement, etc. Based on a proposed clustering algorithm for the training pairs, a new neural network, namely the range-dependent neural network (RDNN) has been developed for better accuracy in hydrological time series prediction. The applicability and potentials of the RDNN in daily streamflow and annual reservoir inflow prediction are examined using data from two watersheds in China. Empirical comparisons of the predictive accuracy, in terms of the model efficiency R 2 and absolute relative errors (ARE), between the RDNN, backpropagation (BP) networks and the threshold auto-regressive (TAR) model are made. The case studies demonstrated that the RDNN network performed significantly better than the BP network, especially for reproducing low-flow events.Key words river flow; prediction; hydrological time series; artificial neural networks; rangedependent neural network; threshold auto-regressive (TAR) model Prévision de séries temporelles de débits en rivière par un réseau de neurones dépendant d'écheDe Résumé Les réseaux de neurones artificiels fournissent une alternative prometteuse pour la modélisation des séries temporelles en hydrologie. Cependant, plusieurs problèmes fondamentaux se posent encore, comme l'identification de la structure, l'estimation des paramètres, la généralisation de l'amélioration des performances, etc. Un nouveau type de réseau de neurones, basé sur un algorithme de classification des couples d'apprentissage et baptisé réseau de neurones dépendant d'échelle, a été développé pour améliorer la précision de la prévision des séries hydrologiques. Nous avons étudié l'applicabilité et le potentiel de ce modèle pour la prévision des débits journaliers et annuels de deux bassins versants chinois. Nous avons comparé empiriquement la qualité de la prévision, en termes d'efficacité du modèle R 2 et d'erreur relative absolue, entre le réseau de neurones dépendant d'échelle, le réseau à rétro-propagation et le modèle autorégressif seuillé. Les études de cas ont montré que le réseau de neurones dépendant d'échelle donne des résultats significativement meilleurs que le réseau à rétro-propagation, en particulier pour la prévision des débits faibles.Mots clefs écoulement en rivière; séries hydrologiques temporelles; réseau de neurones dépendant d'échelle; modèle auto-régressif seuillé

The most popular practice for analysing nonstationarity of flood series is to use a fixed single-type probability distribution incorporated with the time-varying moments. However, the type of probability distribution could be both complex because of distinct flood populations and time-varying under changing environments. To allow the investigation of this complex nature, the time-varying two-component mixture distributions (TTMD) method is proposed in this study by considering the time variations of not only the moments of its component distributions but also the weighting coefficients. Having identified the existence of mixed flood populations based on circular statistics, the proposed TTMD was applied to model the annual maximum flood series of two stations in the Weihe River basin, with the model parameters calibrated by the meta-heuristic maximum likelihood method. The performance of TTMD was evaluated by different diagnostic plots and indexes and compared with stationary single-type distributions, stationary mixture distributions and time-varying single-type distributions. The results highlighted the advantages of TTMD with physically-based covariates for both stations. Besides, the optimal TTMD models were considered to be capable of settling the issue of nonstationarity and capturing the mixed flood populations satisfactorily. The most optimal model with time or physically-based covariates is highlighted in bold. Stationarity in the last column means the situation where distribution parameters do not vary with explanatory variables. 78 L. YAN ET AL.

Non-linearity, uncertainty and subjectivity are the three predominant characteristics of contractors prequalification which lead to the process being more of an art than a scientific evaluation. A fuzzy neural network (FNN) model, amalgamating both the fuzzy set and neural network theories, has been developed aiming to improve the objectiveness of contractor prequalification. Through FNN theory, the fuzzy rules as used by the prequalifiers can be identified and the corresponding membership functions can be transformed. Eightyfive cases with detailed decision criteria and rules for prequalifying Hong Kong civil engineering contractors were collected. These cases were used for training (calibrating) and testing the FNN model. The performance of the FNN model was compared with the original results produced by the prequalifiers and those generated by the general feedforward neural network (GFNN, i.e. a crisp neural network) approach. Contractors' ranking orders, the model efficiency (R2) and the mean absolute percentage error (MAPE) were examined during the testing phase. These results indicate the applicability of the neural network approach for contractor prequalification and the benefits of the FNN model over the GFNN model. The fuzzy neural network is a practical approach for modelling contractor prequalification.Fuzzy Reasoning Neural Network Contractor Prequalification,

The selection criteria for contractor pre‐qualification are characterized by the co‐existence of both quantitative and qualitative data. The qualitative data is non‐linear, uncertain and imprecise. An ideal decision support system for contractor pre‐qualification should have the ability of handling both quantitative and qualitative data, and of mapping the complicated non‐linear relationship of the selection criteria, such that rational and consistent decisions can be made. In this research paper, an artificial neural network model was developed to assist public clients identifying suitable contractors for tendering. The pre‐qualification criteria (variables) were identified for the model. One hundred and twelve real pre‐qualification cases were collected from civil engineering projects in Hong Kong, and 88 hypothetical pre‐qualification cases were also generated according to the ‘If‐then’ rules used by professionals in the pre‐qualification process. The results of the analysis totally comply with current practice (public developers in Hong Kong). Each pre‐qualification case consisted of input ratings for candidate contractors' attributes and their corresponding pre‐qualification decisions. The training of the neural network model was accomplished by using the developed program, in which a conjugate gradient descent algorithm was incorporated for improving the learning performance of the network. Cross‐validation was applied to estimate the generalization errors based on the ‘re‐sampling’ of training pairs. The case studies show that the artificial neural network model is suitable for mapping the complicated non‐linear relationship between contractors' attributes and their corresponding pre‐qualification (disqualification) decisions. The artificial neural network model can be concluded as an ideal alternative for performing the contractor pre‐qualification task.