一种基于BP神经网络的电影商业预测方法(IJITCS-V4-N11-9)

I.J.InformationTechnologyandComputerScience,2012,11,67-73PublishedOnlineOctober2012inMECS()DOI:10.5815/ijitcs.2012.11.09Copyright©2012MECSI.J.InformationTechnologyandComputerScience,2012,11,67-73AMethodofMovieBusinessPredictionUsingBack-propagationNeuralNetworkDebadityaBarmanDepartmentofComputerScienceandEngineeringJadavpurUniversity,Kolkata-700032,IndiaE-mail:[e-mail]rtmentofComputerScienceandEngineeringJadavpurUniversity,Kolkata-700032,IndiaE-mail:[e-mail]—FilmindustryisthemostimportantcomponentofEntertainmentindustry.ProfitandLossbothareveryhighforthisbusiness.Beforereleaseofaparticularmovie,iftheProductionHouseordistributorsgetsanytypeofpredictionthathowthefilmwilldobusiness,thenitcanbehelpfultoreducetherisk.Inthispaperwehaveproposed,backpropagationneuralnetworkforpredictionaboutthebusinessofamovie.Notethat,thismethodissuccessfullyappliedinthefieldofStockMarketPrediction,WeatherPredictionandImageProcessing.IndexTerms—FilmIndustry,ArtificialNeuralNetwork,Back-PropagationI.IntroductionAmovie[1],alsocalledafilmormotionpicture,isaseriesofstillormovingimages.Itisproducedbyrecordingphotographicimageswithcameras,orbycreatingimagesusinganimationtechniquesorvisualeffects.Theprocessoffilmmakinghasdevelopedintoanartformandhascreatedanindustryinitself.Filmsareculturalartifactscreatedbyspecificcultures,whichreflectthosecultures,and,inturn,affectthem.Itisconsideredtobeanimportantartform,asourceofpopularentertainmentandapowerfulmethodforeducatingorindoctrinatingcitizens.Thevisualelementsofcinemagivemotionpicturesauniversalpowerofcommunication.FilmIndustryisanimportantpartofpresent-daymassmediaindustryorentertainmentindustry(alsoinformallyknownasshowbusinessorshowbiz).Thisindustry[2]consistsofthetechnologicalandcommercialinstitutionsoffilmmaking:i.e.filmproductioncompanies,filmstudios,cinematography,filmproduction,screenwriting,pre-production,postproduction,filmfestivals,distribution;andactors,filmdirectorsandotherfilmcrewpersonnel.ThemajorbusinesscentersoffilmmakingareintheUnitedStates,India,HongKongandNigeria.Theaveragecost[3]ofaworldwidereleaseofaHollywoodfilmorAmericanfilm(includingpre-production,filmandpost-production,butexcludingdistributioncosts)isabout$65million.Itcanbestretchedupto$300million[4](PiratesoftheCaribbean:AtWorld'sEnd).Worldwidegrossrevenue[5]canbealmost$2.8billion(Avatar).Profit-lossisfoundtovaryfromaprofit[6]of2975.63%(CityIsland)toaloss[7]of1299.7%(ZyzzyxRoad).SoitwillbeveryusefulifwecandevelopapredictionsystemwhichcanpredictaboutFilm’sbusinesspotential.ManyartificialneuralnetworkbasedmethodsareusedtodesignforsuccessfulStockMarketPrediction[8],WeatherPrediction[9],ImageProcessing[10],andTimeSeriesPrediction[11],andTemperaturePredictionsystem[12]etc.Herewehaveproposedamethodbasedonbackpropagationneuralnetworkforpredictionofprofit/lossofamoviebasedonsomepre-definedgenres.Theformulationoftheproblemispresentedinthenextsection.SectionIIIdescribesourproposedmethod.ThemethodispresentedintheformofanalgorithminsectionIII-A.ExperimentalresultsontenmoviesselectedrandomlyfromagivendatabasecanbefoundinsectionIV.ConcludingremarksandscopeforfurtherworkhasbeenincorporatedinsectionV.II.StatementoftheProblemEveryFilmcanbeidentifiedbycertainfilmgenres.Infilmtheory,genre[13]referstothemethodbasedonsimilaritiesinthenarrativeelementsfromwhichfilmsareconstructed.Mosttheoriesoffilmgenreareborrowedfromliterarygenrecriticism.Somebasicfilmgenresare-action,adventure,animation,biography,comedy,crime,drama,family,fantasy,horror,mystery,romance,science-fiction,thriller,waretc.Onefilmcanbelongtomorethanonegenre.Likethemovietitled“Avatar(2009)”belongsto[14]action,adventure,andfantasygenres.68AMethodofMovieBusinessPredictionUsingBack-propagationNeuralNetworkCopyright©2012MECSI.J.InformationTechnologyandComputerScience,2012,11,67-73Anyfilm’ssuccessishighlydependentonitsfilmgenres.OtherimportantfactorsarereputationofFilmStudioorProductionhouseandpresentpopularityofcastedactor/actress.WecanconsiderthesegenresandthesaidfactorsasaFilm’sattributes.Wecanthencollecttheseattribute’sdataofpastFilms.BasedonthesedatawecanpredictaboutanupcomingFilm’sfuturebusiness.Wehaveused20moviegenreslikeAction,Adventure,Animation,Biography,Comedy,Crime,Documentary,Drama,Family,Fantasy,History,Horror,Musical,Mystery,Romance,ScienceFiction,Sport,Thriller,War,andWestern.Notethatthefactorssuchastheoverallratinggivingbytheviewers,reputationofthefilmdistributorsandpresentpopularityofactor/actressperformedforthefilm,hasbeentakencareofbytheinclusionofthefollowing3attributes-Distributorreputation,overallratingandCastingrating.Alltheseratingaregivenin10-scalerating.Wehaveusedallthese23attribute’sandnormalizedprofitpercentagevaluetotraintheneuralnetwork.Aftertrainingwehaveusedittopredictthenormalizedprofitpercentageofagivenmovie.Artificialneuralnetwork[15]learningmethodsprovidearobustapproachtoapproximatingreal-valued,discrete-valued,andvector-valuedtargetfunctions.Forcertai