Bayesian Classifier with K-Nearest Neighbor Density Estimation for Slope Collapse Prediction

Abstract

Listen

Translate article

Heavy rainfall and typhoon oftentimes cause the collapse of hillslopes across mountain roads. Disastrous consequences of slope collapses necessitate the approach for predicting their occurrences. In practice, slope collapse prediction can be formulated as a deterministic classification problem with two class labels, namely “collapse” and “non-collapse”. Nevertheless, due to the criticality and the uncertainty of the problem, evaluating the collapse susceptibility of an area is a challenging task. This study proposes a novel Artificial Intelligence (AI) approach, named as K-Nearest Neighbor Based Bayesian Classifier (KNNBC), to deal with slope collapse assessment. In the proposed model, Bayesian inference is used as a framework to achieve probabilistic prediction of slope collapse. Meanwhile, K-Nearest Neighbor (K-NN) is employed as a density estimation technique. Equipped with probabilistic outputs, the K-NNBC is able to yield predictions with different levels of confidence and diminish misclassified cases. Experimental results point out that the proposed model is very helpful for decision-makers in slope collapse assessment and disaster prevention planning.