As an environmentally friendly and reliable technology, laser cladding is widely used for surface hardening and damage repair of automotive parts. Combining the trial and error method and Taguchi method, we designed laser melting experiments to investigate the forming process parameters of laser melting 316L alloy powder on the surface of 45# steel. Laser power, scanning speed, and powder feeding speed served as variable parameters, while the dung beetle optimization algorithm with multi strategy fusion was employed to optimize hyper parameters of the kernel limit learning machine. Regression prediction models for evaluation indices of the laser cladding process (width to height ratio and dilution rate) were established, achieving coefficients of determination of 0.973 5 and 0.975 9, with average absolute percentage errors of 0.007 7% and 0.061 2%, respectively. Using Taguchi's method for analyzing the causal analysis, the effects of laser cladding process parameters on the dilution rate were ranked as follows: laser power (P)> powder feeding speed (F)> scanning speed (V), and the effects of aspect ratio were ranked as follows: powder feeding speed (F)> scanning speed (V)> laser power (P). The experimental results show the accuracy of the proposed method, and the ideal regression prediction model of the aspect ratio and dilution rate of cladding layer can be obtained, which can provide a theoretical basis for the quality control of cladding layer.