The user-knowledge crowdsourcing task allocation integrated decision model and genetic matrix factorization algorithm

摘要

Given the pattern matching problem between the knowledge in the Expert Knowledge Recommendation System (EKRS) established by the author of this article and user's expertise knowledge, based on the full consideration of some constraint conditions such as user characteristics, task characteristics and matching eigenvalues, the userknowledge integrated decision-making mathematical model (UKID) is established by using the crowdsourcing task allocation method and matrix decomposition algorithm. Aiming at the characteristics of slow convergence speed and inaccurate user-knowledge (task) matching of the UKID model in a high-dimensional space, the improved genetic matrix factorization algorithm (GMF) is designed. The genetic matrix algorithm is used to map the high-dimensional spatial data to the low-dimensional characteristic space which can improve the accuracy of task assignment. Through analysis to 'hereditary property' under high complexity condition, with the maximum preserved crossover operator and adaptive crossover mutation, GMF not only furthest retains the excellent characteristics of parents and improves the algorithm 'premature properties', but also builds up the quality of the simple genetic algorithm's optimization capabilities and enhances the solution's accuracy. Finally, the algorithm was experimentally verified with Amazon Public dataset and the real data extracted with EKRS. The experimental results showed that the UKID model and algorithm were applicable and could provide the decisionmaking guidance for the knowledge crowdsourcing task assignment.