Predictive Capabilities of Multilayer Perceptron (MLP) in WEKA Algorithm for High Strength Concrete with Steel Fiber Addition

Abstract

In this study, a neural network based model available in Weka Algorithms, was utilized to test the predictive capacity of compressive strength in high strength concrete (HSC) with steel fiber addition. Fiber addition levels ranged from 0.19 – 2.0% were utilized obtained from literature with a total of 192 instances (datasets) and 10 attributes. To test the performance of the algorithm, a 10 – fold cross-validation method was used to assess the effectiveness which was later compared with full training sets. Also, seven learning schemes were utilized to determine the optimum using percentage split. Results generated from the model include correlation coefficient, mean absolute error, root mean squared error, and relative absolute error. It was observed a good correlation coefficient was obtained which was close to unity at 70-30 and 80-20% of training to testing, and significant reduction in the associated errors were observed. Results for coefficient of determination are also presented and follow the same trend observed in the percentage split results. Time taken to generate the model was much shorter an indication of efficiency.