Purpose – The study reported in this paper proposed the use of artificial neural networks (ANN) as viable alternative to regression for predicting the cost of building services elements at the early stage of design. The purpose of this paper is to develop, test and validate ANN models for predicting the costs of electrical services components. Design/Methodology/Approach – The research is based on data mining of over 200 building projects in the office of a medium size electrical contractor. Of the over 200 projects examined, 71 usable data were found and used for the ANN modeling. Regression models were also explored using IBM Statistical Package for Social Sciences Statistics Software 21, for the purpose of comparison with the ANN models. Findings – The findings show that the cost forecasting models based on ANN algorithm are more viable alternative to regression models for predicting the costs of light wiring, power wiring and cable pathways. The ANN prediction errors achieved are 6.4, 4.5 and 4.5 per cent for the three models developed whereas the regression models were insignificant. They did not fit any of the known regression distributions. Practical implications – The validated ANN models were converted to a desktop application (user interface) package – “Intelligent Estimator.” The application is important because it can be used by construction professionals to reliably and quickly forecast the costs of power wiring, light wiring and cable pathways using building variables that are readily available or measurable during design stage, i.e. fully enclosed covered area, unenclosed covered area, internal perimeter length and number of floors. Originality/value – Previous studies have concluded that the methods of estimating the budget for building structure and fabric work are inappropriate for use with mechanical and electrical services. Thus, this study is unique because it applied the ANN modeling technique, for the first time, to cost modeling of electrical services components for building using real world data. The analysis shows that ANN is a better alternative to regression models for predicting cost of services elements because the relationship between cost and the cost drivers are non-linear and distribution types are unknown.

Uncertain water allocations and water trading prices are a key constraint to efficient irrigated cropping and water trading decisions. This study shows that neural network models can reasonably forecast seasonal allocations and trading prices in water markets. These models can complement other forecasting techniques such as regression analysis and time series models as the former can better capture the non-linearities in the water trading system. Using a 50% probability risk factor for water variability, the water allocation model showed minor estimation error; however, in one instance the model underestimated the water allocation by 21%. This may be due to exceptionally low initial water allocations and borrowing of water from future years which was outside the training data sets. Similarly, the water trading price forecast model showed modest estimation error of about 11% during 2004/05 probably due to drought. Overall the models have good water allocation and price forecasting accuracy, and the determinants of water trading prices identified by the neural network models are those expected of the econometric models/economic theory.

Estimation of channel seepage is an essential task in improving the management of earthen channel systems. The spatial distribution of seepage rates along the channels must be quantified to establish the economic and environmental merit of reducing conveyance losses. In Australia, due to recurring droughts and irrigation induced salinity concerns, there is much pressure to improve the efficiency of existing water resource use. Saving seepage losses from earthen channels has therefore become an important issue for several reasons including the loss of a valuable resource, maintaining channel assets and reducing accessions to groundwater.In this paper spatial distribution of channel seepage was quantified using artificial neural networks (ANNs). The electromagnetic imaging (EM31) data along with hydraulic conductivity, depth and salinity of groundwater were correlated with Idaho seepage meter measurements using the ANNs. It is estimated that over 42 million m 3 of water can be lost annually from 500 km of channel in the Murrumbidgee Irrigation Area. The distributed channel seepage analysis indicates that most significant seepage (>20 mm/day) occurs in less than 32% of the surveyed channel length; therefore it is important to target channel lining investments to the leakiest parts -''hotspots'' -of the channel system. Copyright # 2008 John Wiley & Sons, Ltd. RÉ SUMÉ L'évaluation des fuites sur canal est une tâche essentielle pour améliorer la gestion des systèmes de canaux en terre. La distribution spatiale des taux de fuite le long des canaux doit être mesurée pour établir l'intérêt économique et environnemental d'une réduction des pertes de transport. En Australie à cause des sécheresses récurrentes et des inquiétudes sur la salinité induite par l'irrigation, il y a une forte pression pour améliorer l'efficience de l'eau. Ainsi la réduction des fuites sur les canaux en terre devient une question importante pour plusieurs raisons: la perte d'une ressource de haute valeur, la maintenance des investissements et la réduction de l'usage des eaux souterraines.Dans ce papier la distribution spatiale des fuites a été quantifiée en utilisant un réseau neuronal (ANN). Les données électromagnétiques, la conductivité hydraulique, la profondeur et la salinité des eaux souterraines ont été corrélées avec des mesures d'un compteur d'infiltration Idaho en utilisant l'ANN. On estime que plus de 42 millions de mètres cubes peuvent être perdus annuellement sur 500 kilomètres de canal dans le périmètre IRRIGATION AND DRAINAGE

This paper describes an adaptive learning framework for forecasting end-season water allocations using climate forecasts, historic allocation data, and results of other detailed hydrological models. The adaptive learning framework is based on artificial neural network (ANN) method, which can be trained using past data to predict future water allocations. Using this technique, it was possible to develop forecast models for end-irrigation-season water allocations from allocation data available from 1891 to 2005 based on the allocation level at the start of the irrigation season. The model forecasting skill was further improved by the incorporation of a set of correlating clusters of sea surface temperature (SST) and the Southern oscillation index (SOI) data.A key feature of the model is to include a risk factor for the end-season water allocations based on the start of the season water allocation. The interactive ANN model works in a riskmanagement context by providing probability of availability of water for allocation for the prediction month using historic data and/or with the incorporation of SST/SOI information from the previous months.All four developed ANN models (historic data only, SST incorporated, SOI incorporated, SST-SOI incorporated) C opyright c 2010 W iley Perio d icals, In c. 324SEASONAL WATER ALLOCATIONS 325 demonstrated ANN capability of forecasting end-of-season water allocation provided sufficient data on historic allocation are available. SOI incorporated ANN model was the most promising forecasting tool that showed good performance during the field testing of the model.

No abstract