Development Models of Artificial Neural Network and Multiple Linear Regression for Predicting Compression Index and Compression Ratio for Soil Compressibility of Ramadi City

Abstract

Artificial neural networks (ANN) as new techniques employed for the development of predictive models to estimate the needed parameters in geotechnical engineering to be used for comparison with laboratory and field tests and consequently reduce the cost, time, and effort. Flexible computing techniques are using an alternative statistical tool to analyze and evaluate experimental data from 102 consolidation tests on a variety of undisturbed soils from Ramadi city. The regression equations are developed to estimate the compression index and the compression ratio from index data. Multi-Layer Perceptron (MLP) network model is used to calculate compression index and a compression ratio of soils and comparing with the multiple linear regression statistical model MLR. It is found that the MLP showed a higher performance than MLR in predicting Cc and Cr and model accuracy between 0.81 to 16 percent. This will provide a good method for minimizing the potential inconsistency of correlations.