一種融合神經(jīng)與遺傳的食物推薦算法

打開文本圖片集

關(guān)鍵詞：食譜推薦；食物選擇；雙種群遺傳算法；神經(jīng)網(wǎng)絡(luò)；多目標(biāo)優(yōu)化；禁忌搜索算法中圖分類號：TN911.23-34；TP391.3 文獻(xiàn)標(biāo)識碼：A 文章編號：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

隨著經(jīng)濟(jì)全球化和城市化步伐的快速推進(jìn)，社會生活節(jié)奏亦隨之加快。（剩余8495字）