Modeling and parameter ranking of construction labor productivity

For having, both a qualitative as well as a quantitative analysis of various aspects of the topic, a field investigation spanning over eighteen months, for collecting actual data from on-going construction sites was carried out. The set of nine input parameters which is selected for this study is considered to be the one which causes short term or daily variations in productivity. The entire analysis and experimentation can be divided into three main parts, those of modeling the phenomenon, ranking the independent variables in the order of their relative significance and determining and graphically depicting the exclusive influences of each variable on the output which is daily productivity of formwork installation operations. Various neural network models under different paradigms and network settings were developed and compared for performances under the criteria of R 2, MSE and MAE. The best performing model of Back Propagation with modified learning rate and momentum was validated with data unexposed to the model