Abstarct: Abstract Cement-baxsed matrices have low tensile strength and ductility, necessitating improvements in their mechanical properties. The addition of fibers to these matrices can significantly enhance their tensile strength and ductility. Fibers not only improve resistance to cracking but also increase energy absorption capacity and durability compared to fiber-free cement-baxsed matrices. The mechanism of fiber reinforcement occurs through the transfer of stress between the matrix and the fibers, creating mechanical interlocking. Prior to crack formation, tensile stress is distributed between the fibers and the matrix, and after cracking occurs, the total stress is transferred to the fibers. Therefore, understanding the pull-out behavior of fibers is essential for comprehending the tensile and flexural behavior of fiber-reinforced concretes, which can be achieved through pull-out tests and the plotting of corresponding curves. Despite the high validity of pull-out tests, challenges such as production costs of samples, potential errors during testing, and variability in results exist. Thus, employing artificial intelligence methods for predicting fiber pull-out curves appears necessary. This research utilizes four models: CNN1D, CNN2D, LSTM, and XGBoost to predict fiber pull-out curves from cement-baxsed matrices. For model development and training, 502 experimental data samples from laboratory studies were collected. The input parameters used include embedded length in the matrix, angle of inclination of the fibers, type of material and geometric shape of the fibers, end-type of the fibers, pitch of spirals, number of twists in 10 mm, length and diameter of fibers, aspect ratio (length to diameter ratio), loading rate, type and amount of cement per cubic meter, binder amount, silica fume (microsilica), sand, gravel, quartz, superplasticizer, water per cubic meter, water-to-cement ratio (W/C), water-to-binder ratio (W/B), curing age in weeks, and compressive strength (