5.3.1 Netflix Prize

The Netflix Prize amfani bude kira zuwa hango ko hasashen abin da fina-finai da mutane za su so.

A mafi kyau da aka sani bude kira aikin ne Netflix Prize. Netflix sigar online movie haya company, kuma a shekarar 2000 ta kaddamar Cinematch, wani sabis don bayar da shawarar movies to abokan ciniki. Alal misali, Cinematch iya lura cewa ka na son Star Wars da Empire buga Back sa'an nan kuma bayar da shawarar cewa ka duba Return na Jedi. Da farko, Cinematch aiki talauci. Amma, a kan hanya na shekaru masu yawa, Cinematch ci gaba da inganta da ikon hango ko hasashen abin da fina-finai da abokan ciniki zai ji dadin. By 2006, duk da haka, ci gaba a Cinematch plateaued. The masu bincike, a Netflix fitine kyakkyawa da yawa duk abin da za su iya tunanin, amma a lokaci guda, su zargin cewa akwai wasu ideas da zai taimaka ma su, inganta tsarin. Saboda haka, da suka zo sama da abin da yake, a lokacin, a m bayani: an bude kira.

M ga m nasarar da Netflix Prize ya yadda bude kira da aka tsara, kuma wannan zane yana da muhimmanci darussa ga yadda bude kira za a iya amfani da social bincike. Netflix ba kawai sa fitar da wani unstructured request ga ideas, wanda shi ne abin da mutane da yawa tunanin a lõkacin da suka fara duba wani kira a bayyane. Ã'a, Netflix shirya a bayyana matsalar da sauki kimantawa sharudda: sun kalubalanci mutane su yi amfani da wani sa na miliyan 100 movie ratings hango ko hasashen miliyan 3 da aka gudanar da-fita ratings (ratings cewa masu amfani ya yi amma wanda Netflix bai saki). Duk wanda zai iya haifar da wani algorithm da zai iya hango ko hasashen miliyan 3 da aka gudanar da-fita ratings 10% fiye da Cinematch zai lashe 1 dala miliyan. Wannan fili da kuma sauki tambaya kimantawa sharudda-gwada annabta ratings to gudanar da-fita ratings-nufi da cewa Netflix Prize aka firam a cikin irin wannan hanyar da mafita ne sauki don duba fiye da samar. ya juya da kalubale na inganta Cinematch a cikin wani al'amari dace da wani kira a bayyane.

A watan Oktoba na shekara ta 2006, Netflix fito da wani dataset dauke da miliyan 100 movie ratings daga game da game da 500,000 abokan ciniki (za mu yi la'akari da tsare sirri abubuwan da wannan labari release a Babi na 6). The Netflix data za a iya conceptualized matsayin wata babbar matrix da yake kamar 500,000 abokan ciniki da 20,000 movies. A cikin wannan matrix, akwai game da miliyan 100 ratings a kan sikelin daga 1 zuwa 5 taurari (Table 5.2). The kalubale shi ne don amfani da kiyaye data a matrix to hango ko hasashen miliyan 3 da aka gudanar da-fita ratings.

Masu bincike da hackers a duniya da aka kõma zuwa ga kalubale, da kuma ta 2008 fiye da 30,000 da mutane suna aiki a kan shi (Thompson 2008) . A cikin shakka daga cikin hamayya, Netflix samu fiye da 40,000 samarwa mafita daga fiye da 5,000 teams (Netflix 2009) . Babu shakka, ba zai iya Netflix karanta kuma fahimta dukan waɗannan samarwa mafita. Dukan abu gudu smoothly, duk da haka, saboda mafita kasance sauki duba. Netflix iya kawai da kwamfuta kwatanta annabta ratings ga gudanar da-fita ratings da wani pre-kayyade awo (musamman awo da suka kasance sunã ne square tushen da nufin-Squared kuskure). Shi ne wannan ikon da sauri kimanta mafita da ya sa Netflix yarda mafita daga kowa da kowa, wanda ya juya a kira su da muhimmanci, domin mai kyau ideas zo daga wasu m wurare. A gaskiya, da lashe bayani da aka ƙaddamar da wata tawagar fara da uku masu bincike cewa yana da wani kafin kwarewa gini movie shawarwarin tsarin (Bell, Koren, and Volinsky 2010) .

Daya kyau bangare na Netflix Prize shi ne cewa shi ya sa kowa da kowa a duniya to suna da bayani kimanta fairly. A lokacin da mutane uploaded su annabta ratings, ba su bukatar upload da ilimi takardun shaidarka, da shekaru, kabila, jinsi, jima'i fuskantarwa, ko wani abu game da kansu. Saboda haka, da annabta ratings wani sanannen shehun malami Stanford aka bi daidai kamar yadda waɗanda suke daga saurayi a cikin ta gida mai dakuna. Abin takaici, wannan ba gaskiya ba ne, a mafi yawan zamantakewa bincike. Wancan ne, domin mafi yawan zamantakewa bincike, kimantawa sosai lokacin cinyewa kuma partially na ra'ayin wani. Saboda haka, mafi yawan bincike ideas an taba tsanani kimanta, kuma a lõkacin da ra'ayoyin su kimanta, yana da wuya a cire wadanda kimantawa daga mahaliccin ideas. Saboda mafita ne sauki duba, bude kira damar bincike don samun damar dukan yiwuwar m mafita da zai fada ta wurin fasa, idan sun kawai dauke mafita daga sanannun furofesoshi.

Alal misali, a daya batu a lokacin Netflix Prize wani da allon suna Simon Funk posted a kan blog a samarwa bayani dangane da wani mufuradi darajar bazuwar, an m daga mikakke aljabara cewa ba a yi amfani da baya daga sauran mahalarta taron. Funk ta blog post ya lokaci guda fasaha da kuma na yau da kullum weirdly. An wannan blog post kwatanta mai kyau bayani ko kuwa shi wani vata lokaci? A waje na da wani kira a bayyane aikin, da mafita iya taba samu tsanani kimantawa. Hakika Simon Funk ba wani farfesa a kirawo Tech ko MIT, shi ya kasance wani software developer wanda, a lokacin, aka Backpacking kusa New Zealand (Piatetsky 2007) . Idan ya yi i-mel wannan ra'ayin wani injiniya a Netflix, shi kusan lalle dã ba a dauki tsanani.

Abin farin, domin tantancewa sharudda sun bayyana da kuma sauki tambaya, ya annabta ratings aka kimanta, kuma shi ne nan take bayyana a fili cewa tsarin kula sosai m: shi rocketed zuwa na hudu wuri a cikin gasar, mai girma sakamakon da aka ba cewa wasu teams ya riga ya kasance aiki na tsawon watanni a kan matsalar. A ƙarshe, sassan Simon Funk ta m aka yi amfani da kusan dukkan m fafatawa a gasa (Bell, Koren, and Volinsky 2010) .

Gaskiyar cewa Simon Funk zaɓi ya rubuta a blog post bayyana ya kusanta, maimakon kokarin ci gaba da shi asiri, Har ila yau, ya nuna cewa mutane da yawa mahalarta a cikin Netflix Prize ba musamman m da miliyan dollar kyauta. Maimakon haka, mutane da yawa mahalarta kuma da jũna a ji dadin ilimi kalubale da kuma cikin al'umma wanda raya kusa da matsala (Thompson 2008) , ji abin da na sa ran da yawa masu bincike za su iya fahimta.

The Netflix Prize ne mai classic misali wani kira a bayyane. Netflix shirya kai a question tare da wani burin (tsinkaya movie ratings) da kuma nẽme mafita daga mutane da yawa. Netflix ya iya kimanta duk wadannan mafita, domin sun kasance da sauki don tabbatar da ya halicci, da kuma kyakkyawan Netflix tsince mafi kyau bayani. Next, zan nuna maka yadda wannan m za a iya amfani da su a ilmin halitta da kuma doka.