Applying regression analysis to predict and classify construction cycle time

Regression techniques are commonly used for addressing complicated prediction and classification problems in civil engineering thanks to its simplicity. For a given dataset, the linear regression from the input space to the output variables can be achieved by using the "least square error" approach, which minimizes the difference between the predicted and actual outputs. The "least mean square" rule can also be used as a generic approach to deriving solutions onlinear or non-linear regressions. The paper addresses the fundamental algorithms of "least square error" and "least mean square" in order to facilitate the prediction and classification of cycle times of construction operations. The classic XOR problem is selected to verify and validate their performances. A viaduct bridge was installed by launching precast girders with a mobile gantry sitting on two piers. The effectiveness of regression techniques in classifying and forecasting the cycle time of installing one span of viaduct considering the most relevant input factors in connection with operations, logistics and resources are demonstrated.