Tunnels are one of the most widely used engineering structures. The most important criterion for the tunnel stability is the amount of deformations created in it. Squeezing phenomenon is one of the factors causing the tunnels deformation. So far, with the valuable steps taken in this field, there is no comprehensive and correct understanding in fundamental mechanism in squeezing. As a result, in this context, more attention has been paid to development of empirical and semi-analytical methods, which among of them there is only a few limited approach to predict the deformation of tunnel. Therefore, in the present study, it has been tried to estimate of tunnel deformation in squeezing grounds using empirical methods. For this purpose, in first section of this research, comprehensive study initiated in order to collect all researches in this field from 1946 to 2018, along with examination of advantages and disadvantages. Then, data related to the 15 tunneling projects from three different countries of India, Nepal and Bhutan were collected. Subsequent of this section, several methods have been studied, which three methods are chosen for the final analysis. In second section, this data mining and artificial intelligence models which include the Logistic Model Tree (LMT), Model Tree with M5P algorithm and Gene Expression Programming (GEP) was developed using collected database. Then, developed models are evaluated using several criteria. In this study, LMT, GEP and M5P models show the best results on the database, respectively. For example, each of the named models has a determination coefficient (R 2 ) of 89.5, 88.57 and 76.43, respectively. the result obtained from the models shows that generally, the deformation increases by increasing the height of overburden and diameter of tunnel and reducing support stiffness and rock mass quality index. In addition, one of the most important reasons for choosing these methods is that they are presented one or more functions as a result. The final models can be easily used anywhere and for various cases. In other words, these functions are analytical so can be used in most cases. Also, developed models are compared with empirical approaches, which show a good result with very high difference. In addition to the high accuracy of developed models, one of the advantages of this research is to investigate the probabilistic problem that causes uncertainty should be considered and the results are closer to reality.