Indoor localization systems (ILS) based on machine learning (ML) are more effective than previously developed traditional localization methods due to sensitivity to environmental factors, limited robustness and limited adaptability for dynamically changing environments. Instead, these ML-based ILS are simple to deploy in actual scenarios. In the recent past, researchers have proposed novel architectures and methodologies to construct scalable, accurate, reliable, robust, and low-power positioning and tracking systems in response to the severe criteria of these use-cases. Since they provide large area coverage to numerous battery- operated devices, low-power wide-area communication technologies have found a place in the industrial and scientific communities that are focused on the Internet of things. With its long- range capabilities, low power consumption, low data rate and scalability, LoRa technology has become significantly popular for timed applications. This research explores feasibility of using supervised ensemble machine learning techniques including Random Forest, Extra Trees, AdaBoost, Gradient Boosting Machines, XGBoost, Voting, Bayesian post-hoc regularization of random forests, and Evolutionary bagging for robust indoor localization systems while optimizing predictive performance. All the proposed algorithms were tested and evaluated under their different hyperparameters. After extensive analysis and evaluation of selected techniques applied to the dataset, the implementation of an ensemble model using Gradient Boosting and LightGBM, integrated with the VotingClassifier, emerges as the best solution for the ensemble. The tested models provided an approximate accuracy of 91% for the ensemble model, consistent with the individual accuracies of 91% for GradientBoost, 91% for LightGBM, 88% for the Bayesian Post-hoc Regularized Random Forest model, and 87% accuracy in the Evolutionary Bagging approach.