National Kaohsiung First University of Science and Technology

FirstTech Institutional Repository
Not a member yet
    18584 research outputs found

    A Study for Automatic Barcode Image Recognition Methods

    No full text
    [[abstract]]條碼已經廣泛的應用在物流業上,又以商品條碼的EAN-13碼為使用上的最大需求,透過自動化辨識EAN-13碼來獲得產品的資訊,使用影像的辨識方法相較其他辨識方法可以節省成本,且透過相機取得的影像方式也較為靈活,本論文研究在不同的拍攝環境下來進行條碼的條紋解碼和條碼的數字辨識。本論文是利用EAN-13碼的兩個辨識特徵資訊也就是條紋和數字來分開進行解碼以及辨識,最後條碼的辨識結果以條紋為主,數字為第二選擇的辨識結果,方法使用到兩種定位方式以及傾斜校正,在條紋上有使用到五種前處理,數字為三種前處理,最後經由三種二值化來將前景背景分割,條紋解碼方法則是使用傳統的方法,光學字元辨識(OpticalCharacterRecognition,OCR)已發展多年也有不少辨識方法被提出,本論文實驗是將切割後的十三個數字依序使用最近鄰居法(k-NearestNeighborsAlgorithm)、樣板相似度比對、支援向量機(SupportVectorMachine)來進行數字辨識。結合條紋和數字的條碼辨識方法即使遇到前處理無法克服所產生的破損,也可以利用其他條碼特徵資訊來使用,經由實驗本論文所提的方法可以在不同拍攝情況下辨識率達85%,平均每張影像辨識執行時間為1.25秒。[[abstract]]Barcodehasbeenwidelyusedinthelogisticsindustry.TheEAN-13barcodehasverylargedemand,soitisimportantforustoobtaintheproductinformationthroughtheautomaticrecognitionofEAN-13barcode.Theimagerecognitionmethodcansavethecostwhenitiscomparedwithotheridentificationmethods.And,theimagecapturedbythecameraisalsoflexibleinvariousenvironments.Thus,thepurposeofthisthesisistostudythebarcodeimagedecodingandthebarcodedigitnumberrecognition.Inthisthesis,thefeaturesofbaranddigit-numberareseparatelyextractedsuchthatEAN-13coderecognitionsofbaranddigit-numbercanbeindividuallydeveloped.Infinalbarcoderecognitionresults,bardecodingresultisthemajorchoice,theresultofdigit-numberistheminorchoice.Proposedmethodusestwoobjectlocationmethods,onetiltcorrection,fivepre-processingmethodsonthebar,threepre-processingmethodsondigitnumber,andthreebinarizationmethodstosegmenttheforegroundandbackground.Thetraditionalbarcodeconstructionruleisemployedtodecodethebar.Opticalcharacterrecognition(OCR)methodshavebeendevelopedformanyyears,sothek-nearestneighbors(KNN)algorithm,templatesimilaritymatching,andsupportvectormachine(SVM)areusedtorecognizethedigit-numbers.Thecombinationofbaranddigit-numberrecognitionresultscanovercomemanydifficultiesinbarcodeautomaticidentifications,sotheproposedmethodisapromisingapproachtorecognizeEAN-13barcode.Experimentalresultsshowthattheproposedmethodachievestherecognitionrateof85%andtheaveragerecognitiontimeis1.25seconds

    A Study for Automatic Coin Image Recognition Methods

    No full text
    [[abstract]]自動硬幣辨識的領域中,傳統都是利用物理上的方式測量,近年利用視覺辨識廣泛的利用於現代硬幣,也漸漸的發展到古硬幣之中,本論文中提出了一種方法,可以有效的利用視覺辨識現代硬幣與古硬幣,只需要少量的訓練資料就可以得到不錯的辨識率。論文首先會介紹在硬幣辨識上的動機與背景,和古硬幣辨識上的困難與挑戰,需要克服到位移、旋轉、嚴重磨損、先天樣式差異與訓練資料少等因素,因此會介紹本篇論文中使用的三種硬幣特徵向量提取的演算法,局部二值樣式(LocalBinaryPatterns)、梯度方向直方圖(Histogramoforientedgradient)、SIFT(Scale-invariantfeaturetransform),還有其變化型,Dense-SIFT則是使用視覺字袋(Bagofvisualwords)模型做提取與匹配。而硬幣特徵向量比對的方式,利用直方圖比對、泥土搬運(EarthMoverDistance),與支持向量機(SupportVectorMachine),最後會有四個獨立的特徵向量比對結果,透過多重專家投票後決定最終的辨識結果,多重專家能夠有效地提高辨識率,降低錯誤判斷率,如同瞎子摸象般,凝聚各類特徵提取方法的優點,匯集出最後的答案。最終在古硬幣資料庫中可以克服到以上的困難,達到不錯的辨識率,在學習少量的資料後可以應用在古至今各類的硬幣辨識上。[[abstract]]Inthefieldofcoinrecognition,measurementsweretraditionallymadebyphysicalmethods.Recently,computervisionhasbeenwidelyusedintherecognitionofmoderncoins,andalsograduallyappliedtotheareaofancientcoinrecognition.Thisthesisprovidesamethod,whichcaneffectivelyrecognizemodernandancientcoinsbycomputervision,andobtainhighrecognitionratewithsmallamountoftrainingdata.Thisthesiswillfirstlyintroducethemotivationsandbackgroundsofcoinrecognition,aswellasthedifficultiesandchallengesofancientcoinrecognition.Concerningtheproblemoftranslation,rotation,seriouslydamagedcoinsandscarcityoftrainingdata,thisthesiswillintroducethreealgorithmsforcoinsfeaturevectorextraction,includinglocalbinarypatterns,histogramoforientedgradient,SIFT(scale-invariantfeaturetransform)anditsvariation-Dense-SIFT,whichusesthemodelof“Bagofvisualwords”toextractfeatureanddetailedmatching.Astothemethodoffeaturevectormatching,threeapproachesarestudiedincludinghistogrammatching,earthmoverdistanceandsupportvectormachine.Finally,therearefourindividualexperimentalresultsofsinglefeaturevectormatching.And,afinalrecognitionresultwillbedecidedthroughavotemadebymultipleexpertmethod.Multipleexpertmethodnotonlyincreasesrecognitionrateeffectivelybutalsodecreasestherateoffalserecognition.Justlikethe“TheBlindMenandtheElephant”,thismethodcancombinetheadvantagesofvariousfeatureextractionmethods,andobtainsthebetterfinalresult.Eventually,intheexperimentsoftheancientcoindatabase,ithasbeenshownthattheproposedmethodcanovercomevariousdifficultiesofcoinrecognitionandachieveahighrecognitionrate.Afterlearningonlysmallamountofdata,theproposedmethodcanbeappliedtothecoinrecognitionofancientcoinsandmoderncoins

    Simulation and Tracking Test of Driving Behaviors with Gyro

    No full text
    [[abstract]]近年來,智慧運輸系統中的傳感技術發展已經能夠以多種方式通報駕駛人,以提升運輸網路的效率。因為陀螺儀的旋轉軸具有指向固定方向的特點,所以陀螺儀被廣泛的應用在航海與航空的領域上。本論文將陀螺儀安裝於汽車與兩輪車上,通過追蹤陀螺儀的角度,對汽車與二輪車的駕駛行為進行模擬。為了進行模擬,設定幾個駕駛行為的情境與實驗場地。我們分別針對不同情形收集安裝於汽車與二輪車上的陀螺儀的每個旋轉角度數據。最後,卡爾曼濾波器則是設計用於追蹤接收端的角度數據。最後的模擬結果顯示,開發的卡爾曼濾波器可以有效的追蹤到陀螺儀所有角度,也可以看出,透過追蹤所有三個旋轉角度可以預測駕駛行為。因此基於陀螺儀的三個旋轉角度開發安全駕駛輔助系統是可行的,同時也有效率的運用運輸網路。[[abstract]]Recently,thedevelopmentofsensingtechnologiesinintelligenttransportationsystemshasbeenabletoinformvariousvehicleusersinvariouswaysandmakethetransportnetworksmoreeffectively.Duetothefeaturethattherotationaxispointstoafixeddirection,thegyroscopeiswidelyusedinaviationandnavigation.Thisthesiswillperformthesimulationsofdrivingbehaviorsbetweencarsandtwo-wheelvehiclesbytrackingtheanglesoftheinstalledgyroscopesinbothsides.Toperformthesimulations,severalscenariosofdrivingbehaviorsaresetupandtheexperimentalfieldsareselected.Wecollecteachrotationangledataofthegyroscopeincarandtwo-wheelvehicles,respectivelyfordifferentscenarios.Finally,aKalmanisdesignedatthereceivingendfortrackingthecollectedangledataforfurtheranalysis.OursimulationresultsshowthatthedevelopedKalmanfiltercantrackalltheanglesofthegyroscopeeffectively.Italsocanbeseenthatthedrivingbehaviorsarepredictablebytrackingallthethreerotationanglesatthesametime.Hence,developingasafe-drivingassistantsystembasedonthethreerotationanglesofthegyroscopeisfeasible,whichiswellincoordinatingtheuseoftransportnetworks.Keywords:Gyroscope,InternetofVehicle,Kalmanfilter,IntelligentTransportSyste

    Phase Modem Based on Zeros Relocation for Multicarrier Communications

    No full text
    [[abstract]]本論文提出了直接對基頻訊號(Baseband)做分析的架構,抓取基頻訊號的過零點(throughzeropoint),即為訊號在時域(Timedomain)上的根,再用抓取到的根還原代表其訊號的多載波多項式。然而並非所有多載波多項式的零點都可以準確地落在單位圓上,我們使用零點遷移的方式,讓訊號的零點可以全部落在單位圓上,接著再用簡單的微積分定理,就可以還原多載波多項式的係數。由於多載波多項式的實部(I訊號)符合共軛倒數(ConjugateReciprocal)的特性,在還原多項式係數時利用共軛倒數的特性,只需要計算出I訊號的的係數,即可還原完整訊號的多項式係數(符碼)。換句話說,在訊號處理上我們只需要針對I或是Q,就可以將符碼解析回來。在模擬實驗上,將此理論搭配Matlab做針對不同調變與不同載波數量的模擬實驗,證明本論文中提出的方法在處理訊號上可以有很低的錯誤率(BitErrorRate,BER)和向量誤差(ErrorVectorMagnitude,EVM)。最後利用向量訊號產生器以及市售RF模組,再建立一個完整的收發機系統來傳送資料,計算收到的資料錯誤率,同時也證明可以應用在實體介面的傳輸上。[[abstract]]Inthispaper,weproposeanarchitecturethatdirectlyanalyzemulticarrierbasebandsignalforcost-effectivemodulationanddemodulation.Thezerosofamulticarrierbasebandsignalaretherootsofthepolynomialwhosecoefficientsarethesymbols.Weretrievethemulticarrierbasebandsignalaccordingtotherootsofpolynomial.Especiallywhenthezerosallfallontheunitcircle,weeasilyapplyphaseshiftkeying,implementedbyzero-crossingcircuits,fortherootscollectionandbasebandsignalretrievingatthereceiver.However,notallthezerosfallontheunitcircle.Weusezerosrelocationtechniqueatthetransmittertoreformthebasebandsignal(thepolynomial).Thenatthereceiver,weretrievethemulticarriersymbols(polynomialcoefficients)usingtheVieta'sformulas.Moreover,becausethepolynomialrepresentingtheIsignalbecomesconjugatereciprocal,wecanretrieveallthesymbolsbyonlyprocessingtheIsignal.ThissavescircuitareaoftheRFandanalogparts.ThesimulationresultsshowthattheproposedmethodhavemuchlowerBitErrorRate(BER)andsmallerErrorVectorMagnitude(EVM)comparedtothestate-of-the-artfilterbank-basedtransceiverarchitecture.Physically,usingthevectorsignalgeneratorandcommercialsoftware-definedRFmodule,wealsoestablishacompletetransceiversystemtorealizetheproposedarchitectureandtoprovetheperformance

    All-Digital CMOS Digital-to-Time Converter with Improved Resolution

    No full text
    [[abstract]]本篇論文提出具改善解析度之全數位CMOS數位至時間轉換器(DTC),其電路架構包含脈衝產生電路(PulseGenerator,PG)、脈衝擴增電路(Pulse-ExpandingCircuit,PEC)及時間扣抵電路(TimeSubtractor,TS)。此架構之核心電路為PEC,其電路動作採用二進制權重方式進行,將根據輸入數位值(d)來進行脈衝寬度調變,由此實現數位至時間轉換功能;PEC則由數個二對一多工器及脈衝擴增單元所構成。在本篇論文中,將利用改良式電容脈衝擴增單元來達到改善解析度之目的,此方式不僅能改善其解析度,且對於PVT變異也有很好的抑制效果。此外也利用對稱的佈局方式來提高其擴增量之線性度。PG之作用則為產生一額外脈衝當作基底,輸入至PEC來進行脈衝擴增之功能,最後再利用TS扣除由PG產生之脈衝寬度,亦即所產生之脈衝寬度只與輸入數位值有關。本篇論文之實作為四位元全數位CMOS數位至時間轉換器,以TSMC0.35-?mCMOS製程製作,晶片面積為0.037mm2,其模擬解析度為11.5微微秒(ps),積分非線性誤差約於0.3LSB之間,功率消耗在每秒一百萬次之取樣?下為0.14mW。[[abstract]]Anall-digitalCMOSdigital-to-timeconverter(DTC)withimprovedresolutionispresentedinthisdissertation.Thecircuitincludespulsegenerator(PG),pulse-expandingcircuit(PEC)andtimesubtractor(TS).Themainblockinthisarchitecture,PEC,isbasedonbinary-weightedschemewhichisconstitutedby2-to-1MUXsandpulse-expandingunits.Theimproved-capacitancemethodisappliedinthispulse-expandingunittoimproveresolutionandPVTvariation.Inaddition,weimprovedthelinearityofpulse-expandingvaluethroughsymmetricallayout.Theall-digital4-bitCMOSdigital-to-timeconverterwasimplementedin0.35-?mTSMCCMOSprocessandoccupied0.037mm2.Theresolutionissimulatedas11.5ps,integralnonlinearityas0.3LSBandthepowerconsumptionsimulatedatasamplingrateof1MHzwas0.14mW

    Design and Implementation of a Contactless Converter with Asymmetrical Half-Bridge Configurations

    No full text
    [[abstract]]本文探討非接觸式電力傳輸系統,研製一台非接觸式電力轉換器。其主要架構包含非對稱半橋驅動電路、諧振電路、以及穩壓輸出電路。本文首先對非對稱半橋架構動作方式進行說明,分析諧振電路如何幫助驅動電路進行柔性切換,以及諧振電路如何增加初次級側之間的電磁耦合。再者,針對諧振電路其電感元件與電容元件之組合方式,研究計算諧振電路等效電路模型,分析其阻抗與反射阻抗結果,選擇SS(串聯-串聯)諧振電路,並計算電路元件參數。穩壓電路使用具同步整流的升壓電路,額外加入電壓偵測的回授電路。最後,本文實際製作一非接觸式電力轉換器,輸出功率5W,傳輸距離為10mm,可供應小功率電器產品使用,驗證本文所提之電力轉換器的可行性。[[abstract]]Thisthesisistostudythenon-contactpowertransmissionsystem,andpresentsacontactlessconverter.Thearchitectureofthesystemconsistsanasymmetrichalf-bridge(AHB)drivercircuit,resonantcircuit,andaregulatedoutputcircuit.Firstly,theoperationprincipleoftheasymmetrichalf-bridgestructuresispresented,andtheresonantcircuitwasbeenanalyzedtoachievethesoftswitchingfeature.Furthermore,thecoupledmagneticresonancesarechoseninordertoimprovetheelectromagneticfield.Theresonantcircuithasadifferentcombinationofinductanceandcapacitive,anditsequivalentcircuitmodelwasstudiedandanalyzed.ASS(series-series)resonantcircuitwasbeenselectedanddesigned.Inaddition,theregulatorcircuitwasbeendesignedbyaboostconverterwithsynchronousrectifierandafeedbackcircuitforthevoltagedetection.Finally,thispaperimplementedacontactlessconverter,whoseoutputpoweris5wattsthrough10mmairgap;itcanbeverifiedthefeasibilityforapplyingtolow-powerelectricalproducts

    Solving The Open Vehicle Routing Problem By Ant Colony Optimization

    No full text
    [[abstract]]本研究主要探討開放式車輛途程問題(OpenVehicleRoutingProblem,OVRP),OVRP發生在起始點始於場站,且結束在其中一個顧客點的情況,在生活中部份產業的配送作業會聘請第三方物流,從工廠分銷產品至各個客戶點,物流車配送完貨品後便回到物流營業所,以物流公司的立場而言及為一個開放式的路線。在本研究中,假設地圖上有一個商品收集站點,並定義了二種情況之下的開放式車輛途程問題來建構路徑規劃,一是到達集合點後便解散,二為到達集合點後再返回公司。主要是針對司機出發拜訪顧客點之後再至集合點的配送作業,例如,回收廢五金的收集業者將載滿回收物的車輛開到處理廠卸除物品,再空車返回公司。本研究利用蟻群最佳化演算法來解決此問題,以傳統的ACS改良而成AggressiveACO,特性在於以最小化車量數與最小旅行成本以找到最佳化路徑。改良了傳統的搜尋機制,同時派遣多隻螞蟻並同時互相競爭搶奪服務下一個顧客點的機率,使得搜尋變得更激進且具競爭力。並以傳統與改良的求解方法進行測試VRP國際標竿例題,比較二種情況之下分別的二種執行結果,實驗結果顯示本研究之改良方法幾乎能夠解出較優良的解,具有可行性。[[abstract]]ThepurposeofthisstudyistoinvestigatetheOpenVehicleRoutingProblem(OVRP),whichisdifferentthantheconventionalvehicleroutingproblemthatthevehiclesarenotrequiredtoreturntothedepot.ThefocusofthisresearchisslightlydifferentthantheoriginalOVRPinwhichthevehiclesaredepartingfromtheirhomecompanyandendatacollectionsite.Theobjectivesofthisstudyistofindthebestvehicledeploymentofeachhomecompanyandthedispatchingplanwithleastcost.TwocasesofthiskindofOVRParedefinedandstudied,caseoneisthatthevehiclesendatcollectionsiteandcasetwoisthatthevehiclehastobacktothehomecompanydirectlyfromthecollectionsite.Inthisresearch,theAggressiveAntColonyOptimization(AACO)solvingprocedurewhichisinspiredbyAntColonySystemisproposedtosolvetheOVRP.Thissolvingapproachcanyieldbettersolutioncomparingtothetraditionalsearchmechanism.Alltheantsneedtocompetewithoneanothertowinthechancetoservethenextcustomer.ThecomputationalresultsindicatethattheproposedheuristicisabletoefficientlysolveOVRP

    A Construction of Geographic Information System by Graph Database─Using Tokyo Metro System as an Example

    No full text
    [[abstract]]隨著資訊科技日益進步,每天產出的資訊量不斷擴大,為因應日漸龐大的資訊量與多元的資料型態,資料庫相關技術也不斷革新,衍生出NoSQL跳脫傳統的綱要(Schema)框架限制、被賦予更高的彈性。例如,圖形資料庫可以使用節點與節點之間的關係建立資料之間的圖形結構,並提供查詢節點之間的親疏關係以及鄰近節點,或是節點之間最短距離等的相關語法。本研究即是使用圖形資料庫來建置地理資訊系統,將繁複的東京地鐵路線以目前最受歡迎的圖形資料庫軟體Neo4j來處理,並將地鐵路線各車站定點標示於地圖相對位置,以利查詢應用及社群分析應用,降低搭乘東京地鐵的困難,解決乘客在東京地鐵各站轉乘的煩惱。整個研究的目標是:希望藉由東京地鐵地理資訊系統的建置經驗,了解在建置串連台灣鐵路、捷運,甚至公共汽車、藍色公路的交通資訊系統時可能遭遇的問題,進而找到解決的方法,改善台灣整體的交通環境,讓台灣成為智慧城市,提振台灣的整體經濟效益。關鍵詞:NoSQL、圖形資料庫、Neo4j、東京地下鐵路、社群網路。[[abstract]]Withtheescalatingprogressofinformationandcommunicationtechnology,thedailyoutputofinformationisexpanding,inresponsetotheever-increasingamountofinformationanddiversedatatypes,thedatabasetechnologyisalsoinnovative,e.g.,theNoSQLtechnologyalleviatestraditionalapproachbyaschema-lessframework,whichgainsthemeritofhigherflexibility,suchthatthegraphdatabasesofferuserstocreatenodes,togetherwiththeirconnectedrelationships.Besides,userscanissueintuitiveandversatilequerystatementstoretrievetherelationshipsbetweennodes,ortheshortestpathbetweennodes.Thisstudywillemploythemostpopulargraphdatabase,Neo4j,tobuildageographicinformationsystemofTokyoMetronetwork.Weintendtomarkthestationinformationonthemaptoshowtheirrelativepositions,tofacilitatetheapplicationofcommunityanalysisandreducethepassengers’obsession.Theaimofthisstudyistolearnhowthefollow-upcanbeappliedtotherailwaysandMRTsofTaiwanasawhole.Forfutureintegration,weintendtotakethebusandevenblueoceanpathwaysintoaccountinthefuture.ThistestbedplayshopefullyanimportantbaseforthefuturedevelopmentofthesmartcityortheoveralltourismenvironmentinTaiwantoboostoverallgrossnationalprofitandefficiency.Keywords:NoSQL,GraphDatabase,Neo4j,TokyoMetroSystem,SocialAnalysis

    The Study of Phishing Webpage Detection Mechanisms Based on Support Vector Machine and Integrated Feature Extraction Methods

    No full text
    [[abstract]]迎向智慧經濟產業世代,全球市場開始重視FinTech金融科技議題,網路將逐漸取代銀行的部分功能,而網路交易的蓬勃發展將促使企業更加重視資訊安全問題。FinTech目前在臺灣雖然只有開放電子支付業務,但其未來市場不容小覷,相對在資安問題的處理也不容馬虎,尤其網路釣魚攻擊是一種目前常見的網路詐騙手段。隨著網路通訊的發達,釣魚網站的欺騙手段也越來越高明,於偵測釣魚網站攻擊時,其常見的方法為特徵提取。特徵提取是透過數據分析等方式,取得網站重要訊息,找出該網頁的特徵、關鍵字等資訊,並以線性分類器進行分類或預測。本研究首先提出五大特徵提取模組,網址列特徵提取模組用來辨別網址列的偽裝攻擊特徵。網頁異常特徵模組用以判斷網頁與其連結是否在同一網域,以及判斷網頁是否已經失聯。網域特徵提取模組主要是藉由WHOIS數據庫來查詢網頁的網域資訊,若有存活時間過短抑或安全通訊協定的身分偽裝就可能有釣魚網站的可能性。TF-IDF文本挖掘模組結合萊文斯坦距離文字編輯距離模組,主要是用以提取網頁關鍵字,並將提取出的關鍵字匯入搜尋引擎下搜尋,所尋找到的網頁與原始網頁進行萊文斯坦距離(Levenshteindistance)計算,查看其連結是否近似於同一個網域。最後我們將這五大特徵提取模組所提取出的特徵值整合後匯入支援向量機(SVM)來預測及識別釣魚網站。[[abstract]]Towardsthewisdomofeconomicindustrygeneration,globalmarketbegantopayattentiontoFinTech(Financialtechnology).Thenetworktransactionswillgraduallyreplacesomeofthefunctionsinthetraditionalbanks.Thevigorousdevelopmentofonlinetransactionswillenableenterprisestopaymoreattentiontoinformationsecurityissues.FinTechisonlyopentoe-paymentservicescurrentlyinTaiwan.Nevertheless,FinTechcannotbeunderestimatedinfuture.Thehandlingofsecurityissuescannotbesloppysincethephishingattackisacommonmethodofonlinefraud.Today,thephishingattackbecomesmoreandmoresophisticatedwiththeprogressofinternetcommunications.Themajormethodfordetectingthephishingattackisfeatureextractingbyanalyzingtheimportantdataonthewebsitestofindthecharacteristicsofthewebpages,keywordsandotherinformation,andusingalinearclassifiertoclassifyorpredicttheresult.Inthisthesis,wefirstproposefivefeatureextractionmodulesincludingtheaddressbarbasedfeaturesmodule,theabnormalbasedfeaturesmodule,thedomainbasedfeaturesmodule,theTF-IDFtextminingmodule,andtheLevenshteindistanceeditdistancemodule.Theaddressbarbasedfeaturesmoduleisusedtoidentifythecamouflageattackfeatureintheaddressbar.TheAbnormalbasedfeaturesmoduleisusedtodeterminewhetherthewebpageisinthesamedomainasitslink,andwhetherthewebpagehasbeenlost.ThedomainbasedfeaturesmodulequeriesthewebsiteofthedomaininformationthroughtheWHOISdatabase.IfthelifetimeforthedomaininformationistooshortortheSSLidentityiscounterfeit,itmayhavethepossibilityofphishing.TheTF-IDFtextminingmodulecombinedwiththeLevenshteindistanceeditingdistancemoduleisusedtoextractthewebpagekeywords,andsavetheextractedkeywordsintothesearchengineafterperformingtheLevenshteindistancecomputationbythefoundpagesandtheoriginalpagetoseewhetherthelinkissimilartothesamedomain.Finally,wecombinethefeaturevaluesextractedfromthefivefeatureextractionmodulesintothesupportvectormachine(SVM)topredictandidentifythephishingwebpage

    Social Bookmark Collaborative Recommendation Systems based on Word Embedding

    No full text
    [[abstract]]現今網際網路上存在著巨量的資訊,要如何要有效率的閱讀以及管理自己感興趣的知識內容成為一個首要的問題,因此,人們會利用社交書籤網站來管理與分享資訊,雖然社交書籤網站提供很好的管理與搜尋機制,但仍欠缺主動推薦個人化的相似使用者與文章。協同過濾是依據使用者的興趣偏好,先找出與目標使用者興趣相似的使用者,接著分析這些相似使用者所收藏的項目,找出目標使用者可能感興趣的項目並推薦之,為了找出興趣相似的使用者,雖有多種計算相似度的方式,但多數未考量到收藏項目之間的語意關係,例如:傳統的餘弦相似度(Cosinesimilarity)。在以興趣為基礎的社交網路上,若是兩個使用者收藏的項目名稱不相同,但是其項目是屬於類似性質時,本研究稱這種型態的人與人或是人與項目之間的關聯為“隱性”的關係,若使用傳統的計算方式,會因收藏項目名稱不同,相似度計算結果為零,因而找不出這種“隱性”關係的興趣類似之人或項目。本研究設計一套推薦使用者與文章的方法,應用在社交書籤網站上,並以詞嵌入模型(Word2vec)的方式,來計算使用者之間、使用者與文章之間的相似度,過濾與推薦興趣相似的使用者與文章,以Diigo書籤網站的資料集作為實驗資料來源,推薦興趣相似使用者以及從相似使用者所收藏的文章中推薦目標使用者可能感興趣的文章。實驗結果顯示,本研究的方法所推薦的使用者與文章品質良好,皆符合目標使用者的興趣偏好,改善了傳統相似度方法對於計算語意相似度不足的缺點,以及本研究所提出的推薦使用者與文章的方法,均優於熱門推薦與隨機推薦。[[abstract]]Nowadays,hugeinformationontheInternetmakespeoplehardtomanagetheknowledgecontents.Thesocialbookmarkingwebsitecaneasilycollect,manage,andsharetheinformationcontent.Althoughsocialbookmarkingwebsiteprovidesgoodmanagementandsearchingtools,itisstilllackofpersonalizedrecommendersystemstosuggestsimilarusersandarticlesthatthetargetusersmayinterestedin.Collaborativefilteringfindsthesimilaruserwithsamepreferencesbasedonthetargetuser'sinterests,andthensuggestssimilaritemstothetargetuser.Inordertofindsimilarusers,therearenumberofwaystocalculatesimilarityincludingcosinemeasure,butmostofthemdonotconsiderthesemanticrelationshipbetweentwousers’preferencetags,oritemkeywords.Onthesocialnetwork,whentwousershavedifferenttagsoritemkeywords,themeasureoftraditionalcosinesimilaritywillbezero.Thus,therecommendersystemfailstosuggestanysimilarusersoritems.Thisstudyuseswordembeddingmodel(Word2vec)asthesimilaritymeasuretorecommendsimilarusersandarticlestothetargetusersforthesocialbookmarkingwebsite.Theexperimentalplatformis“Diigo”bookmarkingwebsite.Theexperimentalresultsshowthattheproposedmethodcansuccessfullysuggestsimilarusersandarticlestothetargetusers.Ourproposedmethodcanimprovetheperformanceoftraditionalsimilaritycalculation,andoutperformtheperformanceofbothpopularrecommendationandrandomrecommendation

    0

    full texts

    18,584

    metadata records
    Updated in last 30 days.
    FirstTech Institutional Repository
    Access Repository Dashboard
    Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇