Journal of the Faculty of Engineering and Architecture of Gazi University 37:2 (2022) 949-965 Effects of the stochastic and deterministic movements in the optimization processes Ebubekir Seyyarer 1, * , Ali Karcı 2 , Abdullah Ateş 2 1 Department of Computer Programming, Van Yüzüncü Yıl University, Van, 65080, Turkey 2 Department of Computer Engineering, İnönü University, Malatya, 44000, Turkey Highlights: Graphical/Tabular Abstract  Stochastic and deterministic initial populations  Multivariate linear regression model  Obtaining and comparing results with gradient based optimization algorithms Figure A. All modules of the study Purpose: The aim of this study is to show the effect of a deterministic initialization method suggested to be used instead of random start method which is still used in optimization algorithms. Theory and Methods: This study focuses on the topic of initial population in optimization algorithms. Outputs of stochastic and deterministic initial populations are compared. Proposed deterministic initialization method reaches the global optimum much faster, which can be given as an important advantage. Multivariate linear regression (MLR) model is used as the model and iris dataset is used as the dataset. Six gradient descent based optimization methods (Stochastic Gradient Descent (SGD), Momentum, Adagrad, RMSProp, Adadelta ve Adam) are used to minimize the error rate and four error functions (integral of the absolute value of the error (IAE), integral of the time-weighted absolute error (ITAE), Mean Square of the Error (MSE) and integral of the square error (ISE)) are used as objective functions. Results: It is sufficient to run the application with initial population developed using the proposed deterministic method once. Outputs of each application are the same because it is run with fixed values. However, Stochastic initialization methods must be run at least 10 times before the outputs can be presented to literature. It is observed that the outputs of deterministic and stochastic methods are equal. However, in terms of operation time, the deterministic method result in an improvement by 90%. Conclusion: According to deterministic and stochastic initialization methods obtained coefficients and iteration numbers are found to be close. However, temporal gain is achieved from the application that is initialized deterministic. According to comparisons, the linear model obtain using the Adadelta optimization algorithm and the MSE objective function perform best. Keywords:  Deterministic initial population  stochastic initial population  multivariate linear regression  optimization algorithms  iris data set Article Info: Research Article Received: 01.03.2021 Accepted: 29.08.2021 DOI: 10.17341/gazimmfd.887976 Correspondence: Author: Ebubekir Seyyarer e-mail:eseyyarer@yyu.edu.tr phone: +90 432 612 2434