The expansion of human settlements modifies the land use pattern. Changing LULC and rainfall patterns have increased the frequency of landslides in Eastern Bhutan. The Markov Chain Model and deep learning algorithms were used for evaluating the past, present, and future landslide susceptibility in parity with changing LULC and rainfall. To achieve the goal, a total 878 landslide locations from different periods were considered. Landslide susceptibility maps (LSMs) for 1990, 2020, 2050, and 2080 were produced using RF, ANN, and DLNN. For the production of future LSMs in 2050 and 2080, LULC and rainfall had been predicted using the Markov Chain Model and ANN. LSMs were evaluated using the ROC, accuracy, F1 measure, Jaccard, MCC, MRSE, and MAE. All the models worked well, but the DLNN model was the most accurate. In some areas having very high and very low landslide risk, an electric resistivity meter was used to check the resistivity and confirm the susceptibility levels. According to the analysis, if the LULC and rainfall pattern change in the present manner, the area of the very high landslide susceptibility zone will increase by 17.40% in 2080. This work can be helpful to planners and used in other similar areas.