The soil classification is a fundamental step in geotechnical engineering, grouping soils based on similar sets of properties. However, the grouping of soil can be done by different procedures requires financial aid and time-consuming laboratory test and/or in situ tests. The soil classification (soil behavior type) is almost mandatory for all the major construction projects, prior to the design of the respective structure. So, multilayer perception (MLP) techniques can play a major role to overcome the cost and time for the respective construction site to generate the soil classification report. Hence, in this study, MLP models are used to soil behavior type (classify of soil) using cone penetration test (CPT) data. The applied MLP models are artificial neural network (ANN) feedforward backpropagation algorithm. As the input considered for the MLP model are cone penetration test data [tip resistance of the cone (qc), sleeve friction fs, friction ratio (Rf), depth (D)]. As a target, soil behavior type (Ib). The ANN model is validated with performance to see the accuracy of the developed model. Finally, sensitivity analysis was performed, it highlights that the Rf parameter has the most significant influence on predicting the desired output (soil behavior type).