Early detection of stuck pipe incidence in real-time will lead to reducing the downtime and costly corrective actions. Therefore, in the last decades, intensive efforts have been made to develop methods to identify the stuck pipes early symptoms. One of these methods is torque and drag (T&D) modeling. Despite numerous advantages, the modeling-based method has, nowadays, T&D modeling is not common because it requires real-time merge and contextual data of heterogeneous frequency and quality as well as recalibration. The alternative approach is to apply data-driven algorithms to solve real-time torque and drag issues. This approach can be judged to be more robust than the first because real-time drilling data can be used. Besides, the data-driven algorithm does not limit the number of input parameters. From this perspective, this thesis proposes a model based on a machine learning concept focused on real-time detecting the early signs of impending stuck pipe by real-time monitoring the pertinent surface drilling parameters. The newly proposed method uses machine learning linear regression algorithms to predict several parameters such as standpipe pressure, hook load, surface torque. The model can train from the data gained and adjust its performance automatically. The predicted parameters help the model to evaluate the behavior of real-time data. This comparison results in an alert level table for every predicted parameter. The alert level table allows the model to calculate stuck probability and generate an alarm when the probability exceeds the preliminary set limits. As a result, the model is suited to use real-time drilling operation data as input, calculate stuck pipe probability and notify the driller when the risk of stuck pipe is out of threshold and may be recognized as dangerous. The first part of the thesis provides an overview of the problem, the current thesis challenges, and objectives. The main machine learning algorithms and data-based approaches application for early stuck pipe sign detection are denoted in the second part of the thesis. Finally, the third part of the work covers the methodology of the proposed approach in detail. Two case studies of the approach implementation using real-time data are also presented here. In conclusion, the results of the work are presented and summarised. The features and shortcomings of the proposed approach are also discussed. The resulting model proves its ability to implement real-time data monitoring to avoid costly and time-consuming stuck pipe incidents.