modeling the completion time of public school building projects using neural networks

the ministry of education in iraq is confronting a colossal deficiency in school buildings while stakeholders of government funded school buildings projects are experiencing the ill effects of extreme delays caused by many reasons. those stakeholders are particularly worried to know ahead of time (at contract assignment) the expected completion time of any new school building project. as indicated by a previous research conducted by the authors, taking into account the opinions of iraqi experts involved with government funded school building projects, nine major causes of delay in school building projects were affirmed through a questionnaire survey specifically are; the contractor's financial status, delay in interim payments, change orders, the contractor rank, work stoppages, the contract value, experience of the supervising engineers, the contract duration and delay penalty. in this research, two prediction models (a and b) were produced to help the concerned decision makers to foresee the expected completion time of typically designed school building projects having (12) and (18) classes separately. the ann multi-layer feed forward with back-propagation algorithm was utilized to build up the mathematical equations. the created prediction equations demonstrated a high degree of average accuracy of (96.43%) and (96.79%) for schools having (12) and (18) classes, with (r²) for both ann models of (79.60%) and (85.30%) respectively. it was found that the most influential parameters of both models were the ratio of the sum of work stoppages to the contract duration, the ratio of contractor's financial status to the contract value, the ratio of delay penalty to the total value of contract and the ratio of mean interim payments duration to the contract duration.