Optimized Bids Evaluation Model for Improved Performance and Quality Delivery in Public Procurement Sector and Construction Projects

Abstract

Traditionally, public sector procurement and construction project contract awards are largely based on the lowest bid award system. However, this practice has been characterized with problems of inferior quality of construction facilities, high incidence of litigations and frequent cost overruns. Therefore, this study is focused at designing an optimized bids evaluation model to overcome the challenges of the traditional methodologies. A multi-parameter bids responsiveness evaluation model that integrates both the mandatory and weighted subfactors criteria was designed to achieve this purpose. A cross-sectional quantitative and qualitative research technique was employed to formulate the instruments used for the research data collection. Purposive and random sampling techniques were deployed in drawing data samples from respondents to identify cases, make inferences about population, save time and reduce cost of the study. Two hundred and six datasets was collected, 66% of the datasets was devoted to training while the remaining 34% was used for testing during the data modeling. Relative importance index (RII) and ranker’s search method was used to measures the strength of relationship between the observed data and ranking of the relative importance indices of the attributes used respectively. Four different classification algorithms, namely: Pruned Decision Tree (PDT), Logistics Model Tree (LMT), Justified Repeated Incremental Pruning (JRIP) and PART were considered in the modeling. The algorithms were tested to determine the model with the best predictive accuracy. From the experiment, the PDT and JRIP outclassed the other two algorithms in the layer. They both have the same correctly classified instances of 99.4%, mean absolute error of 0.062, true positive rate and false positive rate of 0.994 and 0.001 respectively, the ROC Area of 0.994 and recall weighted average of 0.994 respectively. This proves that both algorithms are suitable for the model. However, the pruned decision tree was preferred the best algorithm as a result of the time taken to build the model. The algorithm took 0.01 seconds compared to JRIP with 0.1 seconds. With this performance, the new model will suitably improve the efficiency of the existing methodologies, guaranteed quality delivery and maximum value in any construction projects. Therefore, the model is highly recommended for efficient bid evaluation in the procurement and construction sectors. Meanwhile, the research still paves the way for future research using additional more inputs, larger database and other background factors.