In the actual production life conditions

a large number of multiple choices can be converted into a multiple-choice knapsack problem (MCKP)

but MCKP is a classic NP-hard problem. Therefore

for very large scale MCKP

it is often only possible to use the particle swarm algorithm

wolf pack algorithm

fish swarm algorithm and so on to solve the problem. For swarm intelligence algorithms

efficient and fast greedy algorithms play a key role in the generation of initial solutions. Based on the convex Pareto algorithm (CPA)

an improved Pareto algorithm (IPA) that can quickly get the linear programming dominated set is proposed. IPA firstly selects the minimum weight item of each set

then computes the value density of all items

and finally chooses the greedy choice of the item according to the value density from high to low. When the value of the greedy option is greater than the original selection of the item set

then IPA is iterated. The simulation results show that compared with CPA

the speed of IPA is increased by 98.86%. The PSO-IPA solution accuracy is increased by an average of 28.92%.