Seite - 181 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

his/herpreference,but it requiresauser to inputa lotof informationsuchas selectionof PoIs (Pointof Interest), hence takesa lotof time toobtainaplan.Tosolve thisproblem, therehavebeenproposedautomatic tourplanning systems (ATPS)whichautomatically select PoIs in the specified area and compose/show users a sightseeing tour including movingwaysbetweenPoIs. InexistingATPS,variousobjective functionsareconsidered.Shiraishi et al.[1]pro- posed an ATPS which solves amulti-objective optimization problem taking into ac- count a trade-off betweenmonetary cost and satisfactiondegree.Wuet al. [2] proposed anATPSwhichmaximizes satisfaction degree taking into account tourist’s remaining staminaasaconstraint. Ingeneral,usersdonotalwaysselect theshortestor thecheapest routeduringsight- seeingtourbecause theirmainpurpose is togetsatisfaction/experience throughsightsee- ingactivities.Whereas, theydonotalwaysmaximize thesatisfactionsince their resource (money, time and stamina) is limited.Hence, for themovebetweenPoIs,wemust con- sider the trade-off between the satisfactionof theuser and the time,moneyand stamina asuser’s resource.For theselectionofPoI,wemustalsocomparemultiplePoIs in terms of the trade-off between the satisfaction and the resource consumption.As such,ATPS can be formulated as amulti-objective optimization problemwith several independent factors. Inthispaper,weformulate thesightseeingtourrecommendationasamulti-objective optimizationproblemwithmoney, timeandstaminaconsumptionduringa tour andsat- isfactiondegreeobtainedby the touras independentvariables.Since thisproblem isNP- hard,we propose a heuristic algorithm to quickly obtain semi-pareto optimal solutions based on genetic algorithmNSGA-II [3].We applied the proposedmethod to planning tours targeting 30 tourist spots in Higashiyama-area of Kyoto, Japan. As a result, we confirmed that our algorithm could output semi-pareto optimal solutions in reasonable time that couldbeused fordecisionmakingunder trade-off. Theremainderof thepaper isstructuredasfollows:Section2overviewstheexisting studies related to our proposal. Section 3 provides the formulation of our target prob- lem. Section 4 describes the proposed algorithm based onGA. Section 5 provides the experimental results toevaluateourmethodandfinallySection6concludes thepaper. 2. RelatedWork Mostof the existing tournavigation systems recommendPoIswithhighaverage review scores and/or according to the user’s preference [4,5,6]. These studies focus on the im- provementofuser satisfactionand recommendasinglePoIusingakindoffilteringsys- tems,but theydonotconsider thewhole tourplanning includingroutesandother factors related to tourism. Some studies support thewhole tour planningby connectingmultiplePoIs.Kurata et al. [7,8]createdaweb-based interactive tourplanningservicecalledCT-Planner.This serviceplansand recommendsa tour routewhileanalyzing theuser’spreference. There are some studies that add other factors related to tourism such asmoney, time and stamina as constraints [9,10,11,12]. For example,Wuet al. [2] considered the staminaasaconstraint.However, theseexistingstudies focusonlyon the satisfactionof users as themain factor and do not consider the trade-off between the satisfaction and other factors suchasmoney, time, andstamina. Y.Hiranoetal. /AMethod forGeneratingMultipleTourRoutes 181