Abstract:
Various studies show that the major cause of delay and cost overrun in developing and
underdeveloped countries is inflation or fluctuation of construction material prices.
Construction materials price fluctuations do not have only a direct effect on project
performance. It can cause a shortage of materials in projects; so the effect is multiplied in
line with the idleness of equipment and manpower. Construction material cost is the
major component of construction project costs, so optimization of procurement costs can
decrease project costs. So, this study proposes a hybrid machine learning model for
constructing material prediction and economical procurement of construction materials to
assist the smooth flow of the construction process. To show how the proposed model
helps in predicting construction material prices and decreasing project costs, the model
was tested in Cement material in the Ethiopian construction industry. The cement price
prediction models developed are Multi-Layer Perceptron (MLP) and Long Short Term
Memory (LSTM) using Python programming language. From historical indicators data including Sep-2010 to May-2022, the Net Foreign Asset of Ethiopia is identified as the
main indicator of cement price fluctuation. The MLP model has better performance than
the LSTM model in cement price prediction. The three and seven-month lag prediction
by the MLP model on the test set has a mean absolute percentage error of 16.01 and 13.67
% but LSTM has 14.9 and 24.8 % respectively. For Procurement optimization, the Deep
Q Network model (DQN) was developed. Using MLP cement price prediction, actual
price, and data from the project as input, the comparison of DQN with the mathematical
EOQ model shows a cost reduction of 76.8%. The price prediction model is helpful toowners and contractors to know the cost of construction in advance and the procurement
model helps contractors to avoid extra costs due to procurement and shortage costs by
optimizing the procurement process considering price fluctuation, construction material
requirement plan, warehouse size, holding, ordering and shortage costs
