TEFERI WORKAFERAHU2026-02-132020-12-16https://etd.hu.edu.et/handle/123456789/762This research presents an Artificial Neural Network based correlation of undrained shear strength and swelling pressure with the standard index test results of Koye-Feche fine grained soil. Simple index test results used for the analysis include unit weight, natural moisture content, liquid limit, plastic limit, liquidity index and plasticity index. Unlike conventional methods, ANNs do not depend on simplified assumptions, have universal function approximation capacity, noisy or missing data resistance, accommodate multiple nonlinear variables for unknown interactions, and have a good generalization capability. A total of four models were proposed for prediction. The first three models were produced for Cu while the remaining one was developed for Ps. Swelling pressure and undrained shear strength were trained in an ANNs program and the results were compared with the experimental values. An automated optimization script has also been developed that can be successfully used to pick the optimal network architecture. Performance indices (i.e. Correlation coefficient and root mean square error) were computed to check the prediction capacity of the ANN models. The devised ANN model was compared with the conventional regression analysis and found superior in all cases. The results obtained from ANN and existing empirical formulas were compared to those obtained from the experiments. It was found that the values predicted from the ANN models match the experimental values much better than those obtained from the equations. Undrained shear strength results showed a strong correlation with the combined Liquid Limit and Plasticity Index input parameters. Whereas, Dry unit weight, natural moisture content, liquid limit and plastic index are powerful predictor of swelling pressure. It has been verified that the ANN models can be used satisfactorily to predict undrained shear strength and swelling pressure as a rapid inexpensive substitute for laboratory techniques. Further research should be conducted to extend all aspects of this research, such as by collecting more data in order to improve resultsenArtificial neural networkSwelling pressureUndrained shear strengthIndex propertyThesis