This study revolves around employing machine learning techniques for assessing the liquefaction potential index (LPI) of soil in earthquakes. Three different techniques: Linear Regression, Ensemble, and DNN have been used. Two linear models Ridge and Lasso, five ensemble learning paradigms namely, AdaBoost, Decision Tree, Gradient Boosting, Random Forest and Extreme Gradient Boost regression and one Deep Neural Network (DNN) which is Convolutional Neural Network (CNN) have been used. Important factors involved in earthquake including cyclic stress ratio, depth of soil stratum, total and effective vertical stresses, peak ground acceleration in terms of g (acceleration due to gravity), earthquake magnitude, SPT-N value (Corrected blow count) and stress reduction factor were used as inputs for estimating the LPI of soils. The liquefaction analysis was done on a dataset comprising 235 cases for training and 59 cases for testing the developed models. Among all the models used in this study the GBR model performed the best which falls under ensemble technique. This study establishes that ensemble models can be relied upon to produce the correct LPI in comparison with Linear Regression and CNN. Furthermore, Local Interpretable Model Agnostic Explanation (LIME) has been used to check the influence of each parameter and it is observed that two features CSR and (N1)60 are mostly affecting the prediction.