In order to solve the problems of insufficient accuracy in predicting the scale of fractures in tight reservoirs, a multi-objective regression prediction model was constructed by integrating the black winged kite algorithm (BKA) and random forest (RF) algorithm. Firstly, taking the Chang 4+5 and Chang 6 reservoirs in the Ordos Basin as the research objects, a large-scale sample set was generated through on-site data and FrSmart fracturing numerical simulation software, covering various geological and construction conditions; Secondly, Pearson correlation coefficient and random forest algorithm were used to determine the key geological and construction factors that affect fracture length, width, and height, and to conduct correlation analysis and importance ranking of these factors; Finally, the black winged kite algorithm was used to optimize the hyper parameters of the random forest model and predict the fracture scale characteristics. The results indicate that construction parameters have the most significant impact on fracture scale, while geological parameters mainly control fracture morphology. The constructed BKA-RF model outperforms particle swarm optimization-random forest (PSO-RF） in predicting fracture length, width, and height. The average relative error of the test set for fracture length prediction is only 2.44%, and the coefficient of determination R² exceeds 0.94. This model not only provides reliable support for optimizing fracturing parameters and on-site construction design, but also offers a new technological path for efficient development of tight oil and gas reservoirs.