Abstract: The essence of the first category of assembly line balancing problem is to find the optimal assembly line layout with a given cycle time as a constraint. Traditional global search solving methods are computationally complex and time-consuming. In response to the practical application of customers only considering the optimal cost without considering the assembly line layout scheme, the search problem is transformed into a prediction problem, and the use of artificial intelligence algorithms to predict the optimal cost is proposed. The bottleneck of the key factors that constrain the optimal cost cannot be quickly found using feature importance ranking analysis. Firstly, in the case of changes in worker costs, a new simulation dataset is constructed by solving an integer linear programming model; Secondly, using the above data, random forest regression, decision tree, and XGBoost algorithm models were trained, and optimal cost prediction was achieved using seven parameters including the first type of worker cost and the second type of worker cost; Finally, three different methods were used to rank the importance of features in order to identify the key factors that constrain the optimal cost. The comprehensive performance of three regression algorithms was evaluated using four indicators: R², RMSLE, EV, and MPE. It was found that the XGBoost algorithm had the highest MPE error of 5.12%, while the random forest regression algorithm had the lowest MPE error of 4.09%, proving the feasibility of using intelligent algorithms to predict the optimal cost. A new method for predicting the optimal cost using intelligent algorithms is provided, and the results of feature importance ranking show that the impact of the first type of worker's working time on the optimal cost is significantly higher than other factors.