Prediction of Squeezing Potential in Tunneling Projects Using Data Mining-Based Techniques

Prediction of squeezing conditions is an outstanding issue for appropriate support design and suitable construction procedures in tunneling projects. In this paper, two data mining-based models are presented for prediction of tunnel squeezing. In one of these models, the k-nearest neighbor (k-NN) classifier is used and in the other one the C5.0 decision tree classifier is employed. A database including 115 case histories from different tunneling projects in Himalayan region is applied for development of the models. These models predict the squeezing conditions based on three input parameters, i.e. overburden depth (H), tunnel span or diameter (B), and rock mass quality (Q). The evaluation of proposed models illustrates that both models have good performances, but the k-NN model is a little better than C5.0 model. Furthermore, these models have higher prediction capacities than the common empirical equations. Finally, the results indicate that the H is the most effective input parameter for squeezing perdition based on both k-NN and C5.0 models.