Traditional and ensemble methods are linear models which are considered the most popular techniques for various learning tasks for the prediction of both nominal and numerical values. In this study, we demonstrate the novel concept and working of an algorithm, which customizes the idea of various classification problems with the use of logistic regression in place of linear regression, called a Logistic Model Tree (LMT) algorithm. This study briefly describes the analytical and mathematical implementation of LMT on geographical data for the prediction of rainfall. A step-wise approach is used for the construction of a LMT, which involves a decision tree inducer (C4.5) for the splitting criteria and logistic regression functions for the pruning in which standard regression errors using Cost-Complexity Pruning (CCP) are calculated at each node. This work assesses the abilities of the LMT for the prediction of rainfall across the Kashmir province of the Union Territory of Jammu & Kashmir, India. The implementation methodology was prepared based on six years of historical-geographical data of Kashmir province. It was collected from three different substations having four explanatory independent variables, namely: max temp, min temp and humidity measured at 12 A.M and 3 P.M, moreover a target variable indicating presence and absence of rain. The overall result shows that LMT performs better with the accuracy of 87.23%. At the later stage, we compared the performance of LMT to several algorithms on the same set of data, and show that LMT produces more accurate and compact results.