一种融合神经与遗传的食物推荐算法

打开文本图片集

关键词：食谱推荐；食物选择；双种群遗传算法；神经网络；多目标优化；禁忌搜索算法中图分类号：TN911.23-34；TP391.3 文献标识码：A 文章编号：1004-373X（2025）10-0173-06

Abstract：Peoplepaymoreandmoreatentiontothenutritionandbalanceoftheirdiet，thedemand forfood choices is also higher.Inalusiontotheproblemsoflackofnutrientbalance，lackof diversityandtime-consumingformulationofexisting recipes，adoublepopulationneuralnetwork-geneticalgorithm（DDNT-GA）algorithmisconstructedbyfusingneural networksand combiningdual-populationgeneticalgorithmNSGA-Itogeneratespecificrecipes.Inthisalgorithm，theneuralnetwork isused tolowerthefitnessofoverlyfitindividualstoeffectivelypreventflingintolocaloptima.Individualswithlowfitessare removedtoformanelitestrategy，screenoutthemostsuitable individuals，andimprovetheeficiencyofthemodelwhile achievingfoodnutritionbalance.Byoptimizingtheneuralnetworkandintroducing theregularizationDropoutstrategy，the trainingspeedisimproved.ByusingtheimprovedNSGA-Igeneticalgorithmandincorporatingthedual-populationidea，the taboosearchalgorithmisusedinthesub-populationtoprevent thegenerationofsimilarrecipesbymeansofthetaboolist，soas torealizetherecipediversification.Theexperimentalresultsshowthat，incomparisonwithdepgeneticalgorithms（GA-D，BPGA，NT-GA，JANUS），DDN-GAalgoritmcanincreasethefitnssby11.3%andshortenthetrainingtie.Theresultingipe notonlyhasdiversechangesinfoodombinationsbutalsoimprovestheefiiencyofselectingrcipes，andhasertainpractical value in consumer recipe formulation.

Keywors：reciperecommendation；foodselection;dual-populationgeneticalgorithm；neural network；multi-objective optimization; taboo search algorithm

0 引言

随着经济全球化和城市化步伐的快速推进，社会生活节奏亦随之加快。（剩余8495字）