The choice of optimizer highly decides the performance efficiency of machine learning models particularly within the purview of predictive modelling for regression activities. Thus, this study provides an extended comparative evaluation of various well-known optimization algorithms that can be placed broadly under adaptive moment estimation methods. We pay special attention to the difficult problem of the house price prediction and benchmark performance of four widely applied optimizers: SGD, AdaGrad, RMSProp, and Adam, in terms of their convergence speed, stability, and the possibility of learning huge datasets. Finally, we carry out thorough experiments to investigate under which conditions algorithmic properties of these optimizers correspond to the success of regression models and outline the strengths and weaknesses of each. The algorithm with the best rate among those tested for the problem, Adam showed faster convergence and noise immunity. However, the challenges that the algorithm faces in particular should be specially stated, such as its extreme sensitivity to the chosen configurations of hyperparameters and the possibility of encountering convergence problems when the problem is non-convex. Results of the carried-out work give significant guidance to practitioners who strive to improve regression models inside practical applications of predictions.