{"id":53531,"date":"2019-12-04T00:00:00","date_gmt":"2019-12-03T21:00:00","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/ishhem-anomalii-i-predskazyvaem-sboi-s-pomoshhyu-nejrosetej"},"modified":"2020-02-18T14:01:26","modified_gmt":"2020-02-18T11:01:26","slug":"ishhem-anomalii-i-predskazyvaem-sboi-s-pomoshhyu-nejrosetej","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/administrirovanie\/ishhem-anomalii-i-predskazyvaem-sboi-s-pomoshhyu-nejrosetej","title":{"rendered":"Gjejm\u00eb anomalit\u00eb dhe parashikojm\u00eb d\u00ebshtimet me ndihm\u00ebn e rrjeteve neuronale","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"Gjejm\u00eb<\/p>\n

<\/p>\n

Zhvillimi industrial i sistemeve softuerike k\u00ebrkon v\u00ebmendje t\u00eb madhe ndaj q\u00ebndrueshm\u00ebris\u00eb s\u00eb produktit p\u00ebrfundimtar, si dhe reagim t\u00eb shpejt\u00eb ndaj d\u00ebshtimeve dhe problemeve, n\u00ebse ato ndodhin. Monitorimi sigurisht ndihmon n\u00eb reagimin ndaj d\u00ebshtimeve dhe problemeve m\u00eb efikas dhe m\u00eb t\u00eb shpejt\u00eb, por nuk \u00ebsht\u00eb mjaftuesh\u00ebm. S\u00eb pari, \u00ebsht\u00eb shum\u00eb e v\u00ebshtir\u00eb t\u00eb kontrollohet nj\u00eb num\u00ebr i madh server\u00ebsh \u2013 nevojitet nj\u00eb num\u00ebr i madh njer\u00ebzish. S\u00eb dyti, \u00ebsht\u00eb e nevojshme t\u00eb kuptosh mir\u00eb se si \u00ebsht\u00eb nd\u00ebrtuar aplikacioni, p\u00ebr t\u00eb parashikuar gjendjen e tij. Prandaj, nevojiten shum\u00eb njer\u00ebz q\u00eb kuptojn\u00eb mir\u00eb sistemet q\u00eb po zhvillojm\u00eb, treguesit dhe ve\u00e7orit\u00eb e tyre. Le t\u00eb supozojm\u00eb se madje n\u00ebse gjenden mjaft njer\u00ebz q\u00eb duan t\u00eb merren me k\u00ebt\u00eb, k\u00ebrkohet aq koh\u00eb p\u00ebr t'i trajnuar ato.<\/p>\n

<\/p>\n

\u00c7far\u00eb t\u00eb b\u00ebjm\u00eb? K\u00ebtu na vjen n\u00eb ndihm\u00eb Inteligjenca Artificiale. Artikulli do t\u00eb flas\u00eb p\u00ebr mir\u00ebmbajtjen parashikuese<\/a><\/noindex> (predictive maintenance). Ky qasje po fiton vazhdimisht popullaritet. Jan\u00eb shkruar shum\u00eb artikuj, duke p\u00ebrfshir\u00eb edhe n\u00eb Habra. Kompanit\u00eb e m\u00ebdha po e p\u00ebrdorin k\u00ebt\u00eb qasje p\u00ebr t\u00eb mb\u00ebshtetur funksionimin e server\u00ebve t\u00eb tyre. Pas studimit t\u00eb nj\u00eb numri t\u00eb madh artikujsh, ne vendos\u00ebm ta provojm\u00eb k\u00ebt\u00eb qasje. \u00c7far\u00eb doli nga kjo? <\/p>\n

<\/a><\/noindex><\/p>\n

Hyrje<\/h2>\n

<\/p>\n

Sistemi softuerik t\u00eb zhvilluar her\u00ebt ose von\u00eb del n\u00eb p\u00ebrdorim. \u00cbsht\u00eb e r\u00ebnd\u00ebsishme p\u00ebr p\u00ebrdoruesin q\u00eb sistemi t\u00eb funksionoj\u00eb pa prishje. N\u00ebse ndodhin situata t\u00eb jasht\u00ebzakonshme, ato duhet t\u00eb eliminohen me vonesa minimale. <\/p>\n

<\/p>\n

P\u00ebr t\u00eb thjeshtuar mb\u00ebshtetje teknike t\u00eb sistemit softuerik, ve\u00e7an\u00ebrisht n\u00ebse ka shum\u00eb server\u00eb, zakonisht p\u00ebrdoren programe monitorimi, t\u00eb cilat marrin metrika nga sistemi softuerik q\u00eb punon, ofrojn\u00eb mund\u00ebsi p\u00ebr t\u00eb diagnostikuar gjendjen e saj dhe ndihmojn\u00eb n\u00eb p\u00ebrcaktimin e asaj q\u00eb shkaktoi prishjen. Ky proces quhet monitorimi i sistemit softuerik.<\/p>\n

\n\"Gjejm\u00eb<\/p>\n

Figura 1. Nd\u00ebrfaqja p\u00ebr monitorimin grafana<\/em><\/p>\n

Metodat jan\u00eb tregues t\u00eb ndrysh\u00ebm t\u00eb nj\u00eb sistemi programor, mjedisi t\u00eb ekzekutimit t\u00eb tij ose t\u00eb nj\u00eb makine p\u00ebrpunuese fizike, n\u00ebn t\u00eb cil\u00ebn funksionon sistemi me nj\u00eb etiket\u00eb kohe, n\u00eb momentin kur jan\u00eb marr\u00eb metodat. N\u00eb analiz\u00ebn statike, t\u00eb dh\u00ebnat e metrikave quhen serira temporale. P\u00ebr t\u00eb v\u00ebzhguar gjendjen e sistemit programor, metrikat paraqiten n\u00eb form\u00eb graforesh: n\u00eb akset X \u2013 koha, dhe n\u00eb akset Y \u2013 vlerat (figura 1). Nj\u00eb sistem programor n\u00eb funksion mund t\u00eb regjistroj\u00eb disa mij\u00ebra metrika (nga \u00e7do nyj\u00eb). Ato formojn\u00eb nj\u00eb hap\u00ebsir\u00eb t\u00eb metrikave (serira temporale shum\u00ebdimensionale). <\/p>\n

<\/p>\n

P\u00ebr shkak se sistemet programore komplekse regjistrojn\u00eb nj\u00eb num\u00ebr t\u00eb madh metrikash, monitorimi manual b\u00ebhet nj\u00eb detyr\u00eb e komplikuar. P\u00ebr t\u00eb reduktuar sasin\u00eb e t\u00eb dh\u00ebnave t\u00eb analizuar nga administratori, mjetet e monitorimit p\u00ebrmbajn\u00eb mjete p\u00ebr identifikimin automatik t\u00eb problemeve t\u00eb mundshme. P\u00ebr shembull, mund t\u00eb konfigurohet nj\u00eb tregues q\u00eb aktivizohet n\u00eb rast se hap\u00ebsira e lir\u00eb n\u00eb disk zvog\u00eblohet n\u00ebn nj\u00eb prag t\u00eb caktuar. Gjithashtu, mund t\u00eb diagnostikohet automatikisht ndalimi i serverit ose ngadal\u00ebsimi kritik i sh\u00ebrbimit. N\u00eb praktik\u00eb, mjetet e monitorimit munden mir\u00eb me zbulimin e d\u00ebshtimeve t\u00eb ndodhura tashm\u00eb ose identifikimin e simptomave t\u00eb thjeshta t\u00eb d\u00ebshtimeve t\u00eb ardhshme, por n\u00eb p\u00ebrgjith\u00ebsi parashikimi i d\u00ebshtimeve t\u00eb mundshme mbetet nj\u00eb sfid\u00eb p\u00ebr ta. Parashikimi p\u00ebrmes analiz\u00ebs manuale t\u00eb metrikave k\u00ebrkon angazhimin e specialist\u00ebve t\u00eb kualifikuar. Ky proces \u00ebsht\u00eb me prodhueshm\u00ebri t\u00eb ul\u00ebt. Shumica e d\u00ebshtimeve t\u00eb mundshme mund t\u00eb mbeten t\u00eb paivestuara.<\/p>\n

<\/p>\n

S\u00eb fundmi, mes kompanive t\u00eb m\u00ebdha IT q\u00eb zhvillojn\u00eb software ka pasur nj\u00eb rritje t\u00eb njohjes p\u00ebr at\u00eb q\u00eb quhet sh\u00ebrbim parashikues i sistemeve programore. Thelbi i k\u00ebtij qasje \u00ebsht\u00eb identifikimi i defekteve q\u00eb \u00e7ojn\u00eb n\u00eb degradimin e sistemit n\u00eb faza t\u00eb hershme, para se ai t\u00eb d\u00ebshtoj\u00eb, duke p\u00ebrdorur inteligjenc\u00ebn artificiale. Ky qasje nuk p\u00ebrjashton plot\u00ebsisht monitorimin manual t\u00eb sistemit. Ai \u00ebsht\u00eb ndihm\u00ebs p\u00ebr procesin e monitorimit n\u00eb p\u00ebrgjith\u00ebsi. <\/p>\n

<\/p>\n

Instrumenti kryesor p\u00ebr realizimin e sh\u00ebrbimit parashikues \u00ebsht\u00eb detyra e gjetjes s\u00eb anomalis\u00eb n\u00eb serit\u00eb temporale, pasi n\u00eb rast t\u00eb shfaqjes s\u00eb nj\u00eb anomalie<\/strong> n\u00eb t\u00eb dh\u00ebna ka nj\u00eb probabilitet t\u00eb lart\u00eb q\u00eb pas pak koh\u00ebsh do t\u00eb ndodhi nj\u00eb d\u00ebshtim ose defekt<\/strong>. Anomalit\u00eb jan\u00eb disa devijime t\u00eb treguesve t\u00eb sistemit programor, si\u00e7 \u00ebsht\u00eb identifikimi i degradimit t\u00eb shpejt\u00ebsis\u00eb n\u00eb ekzekutimin e nj\u00eb lloj k\u00ebrkese ose ulja e numrit mesatar t\u00eb k\u00ebrkesave t\u00eb trajtuara n\u00eb nj\u00eb nivel t\u00eb caktuar t\u00eb seancave t\u00eb klient\u00ebve.<\/p>\n

<\/p>\n

Detyra e gjetjes s\u00eb anomali n\u00eb sistemet programore ka specifik\u00ebn e saj. N\u00eb parim, p\u00ebr \u00e7do sistem programor duhet zhvilluar ose p\u00ebrmir\u00ebsuar metodat ekzistuese, sepse gjetja e anomali varet shum\u00eb nga t\u00eb dh\u00ebnat, n\u00eb t\u00eb cilat kryhet, dhe t\u00eb dh\u00ebnat e sistemeve programore ndryshojn\u00eb shum\u00eb var\u00ebsisht nga mjetet e implementimit t\u00eb sistemit deri tek makina kompjuterike n\u00ebn t\u00eb cil\u00ebn \u00ebsht\u00eb ekzekutuar.<\/p>\n

<\/p>\n

Metodat e gjetjes s\u00eb anomali n\u00eb parashikimin e d\u00ebshtimeve t\u00eb sistemeve programore<\/h2>\n

<\/p>\n

N\u00eb radh\u00eb t\u00eb par\u00eb, duhet t\u00eb thuhet se ideja e parashikimit t\u00eb d\u00ebshtimeve \u00ebsht\u00eb frym\u00ebzuar nga artikulli \u00abM\u00ebsimi i makinerive n\u00eb IT-monitorim\u00bb<\/a><\/noindex>. P\u00ebr t\u00eb provuar efektivitetin e qasjes me k\u00ebrkimin automatik t\u00eb anomali, u zgjodh sistemi programor \u00abWeb-Konsolidimi\u00bb, i cili \u00ebsht\u00eb nj\u00eb nga projektet e kompanis\u00eb NPO \u00abKrista\u00bb. P\u00ebr t\u00eb, m\u00eb par\u00eb \u00ebsht\u00eb kryer monitorim manual i metrikave t\u00eb marra. Duke qen\u00eb se sistemi \u00ebsht\u00eb mjaft i nd\u00ebrlikuar, p\u00ebr t\u00eb merret nj\u00eb num\u00ebr i madh metrikash: treguesit e JVM (ngarkesa e mbledh\u00ebsit t\u00eb plehrave), treguesit e OS-s\u00eb n\u00ebn t\u00eb cil\u00ebn ekzekutohet kodi (memoria virtuale, % ngarkesa e CPU-s\u00eb), treguesit e rrjetit (ngarkesa e rrjetit), t\u00eb serverit vet\u00eb (ngarkesa e CPU-s\u00eb, memories), metrikat e wildfly dhe metrikat e veta t\u00eb aplikacionit p\u00ebr t\u00eb gjitha subsistemet kritike. <\/p>\n

<\/p>\n

T\u00eb gjitha metrikat merren nga sistemi me ndihm\u00ebn e graphite. Fillimisht u p\u00ebrdor baza whisper si zgjidhje standarde p\u00ebr grafan\u00eb, por me rritjen e baz\u00ebs s\u00eb klient\u00ebve, graphite nuk p\u00ebrballoi m\u00eb, duke shteruar kapacitetin e sistemit disk t\u00eb DC. Pas k\u00ebsaj, u mor vendimi p\u00ebr gjetjen e nj\u00eb zgjidhjeje m\u00eb efikase. Zgjedhja u b\u00eb p\u00ebr graphite+clickhouse<\/a><\/noindex>, e cila leht\u00ebsoi ndjesh\u00ebm ngarkes\u00ebn mbi sistemin disk dhe e reduktoi n\u00eb pes\u00eb deri n\u00eb gjasht\u00eb her\u00eb hap\u00ebsir\u00ebn e p\u00ebrdorur n\u00eb disk. M\u00eb posht\u00eb \u00ebsht\u00eb paraqitur skema e mekanizmit t\u00eb mbledhjes s\u00eb metrikave me p\u00ebrdorimin e graphite+clickhouse (figur\u00eb 2).<\/p>\n

\n\"Gjejm\u00eb
\n<\/p>\n

Figura 2. Skema e marrjes s\u00eb metrikave<\/em><\/p>\n

Skema \u00ebsht\u00eb marr\u00eb nga dokumentacioni internal. Ajo tregon shk\u00ebmbimin e t\u00eb dh\u00ebnave midis grafana (nd\u00ebrfaqja e p\u00ebrdoruesit p\u00ebr monitorim, q\u00eb ne p\u00ebrdorim) dhe graphite. Marrja e metrikeve nga aplikacioni realizohet nga nj\u00eb software i ve\u00e7ant\u00eb \u2013 jmxtrans<\/a><\/noindex>. Ai gjithashtu i ruan ato n\u00eb graphite.
\nSistemi 'Web-Konsolidim' ka nj\u00eb s\u00ebr\u00eb ve\u00e7orish q\u00eb krijojn\u00eb probleme p\u00ebr parashikimin e d\u00ebshtimeve:<\/p>\n

<\/p>\n

    \n
  1. shpesh ndodh ndryshimi i trendit. P\u00ebr k\u00ebt\u00eb sistem programor l\u00ebshohen versione t\u00eb ndryshme. \u00c7do nj\u00ebra nga ato sjell ndryshime n\u00eb pjes\u00ebn programore t\u00eb sistemit. N\u00eb p\u00ebrputhje me k\u00ebt\u00eb, zhvilluesit ndikojn\u00eb drejtp\u00ebrdrejt n\u00eb metrikat e k\u00ebtij sistemi dhe mund t\u00eb shkaktojn\u00eb ndryshimin e trendit; <\/li>\n
  2. ve\u00e7oria e realizimit, si dhe q\u00ebllimet e p\u00ebrdoruesve t\u00eb k\u00ebtij sistemi shpesh shkaktojn\u00eb anomali pa degrada t\u00eb m\u00ebparshme; <\/li>\n
  3. p\u00ebrqindja e anomalive n\u00eb raport me t\u00eb gjith\u00eb grupin e t\u00eb dh\u00ebnave \u00ebsht\u00eb e vog\u00ebl (< 5%); <\/li>\n
  4. mund t\u00eb ndodhin nd\u00ebrprerje n\u00eb marrjen e treguesve nga sistemi. N\u00eb disa periudha t\u00eb shkurt\u00ebr kohore, sistemi i monitorimit nuk arrin t\u00eb marr\u00eb metrikat. P\u00ebr shembull, n\u00ebse serveri \u00ebsht\u00eb i ngarkuar. Kjo \u00ebsht\u00eb kritike p\u00ebr trajnimet e rrjeteve nervore. Paraqitet nevoja p\u00ebr t\u00eb plot\u00ebsuar boshll\u00ebqet sintetikisht;<\/li>\n
  5. Rastet me anomali shpesh jan\u00eb t\u00eb r\u00ebnd\u00ebsishme vet\u00ebm p\u00ebr nj\u00eb num\u00ebr\/muaj\/kohe specifike (sezonalisht). Ky sistem ka nj\u00eb rregull t\u00eb qart\u00eb p\u00ebrdorimi nga p\u00ebrdoruesit e tij. P\u00ebr pasoj\u00eb, metrikat jan\u00eb t\u00eb r\u00ebnd\u00ebsishme vet\u00ebm p\u00ebr koh\u00eb specifike. Sistemi mund t\u00eb p\u00ebrdoret jo vazhdimisht, por vet\u00ebm n\u00eb disa muaj: selektivisht n\u00eb var\u00ebsi t\u00eb vitit. Ndodhin situata kur e nj\u00ebjta sjellje e metrikave n\u00eb nj\u00eb rast mund t\u00eb \u00e7oj\u00eb n\u00eb d\u00ebshtimin e sistemit programor, kur n\u00eb nj\u00eb rast tjet\u00ebr nuk ndodh.
    \nNga fillimi jan\u00eb analizuar metodat e zbuluar t\u00eb anomalive n\u00eb t\u00eb dh\u00ebnat e monitorimit t\u00eb sistemeve programore. N\u00eb artikujt mbi k\u00ebt\u00eb tem\u00eb, me p\u00ebrqindje t\u00eb vogla anomali n\u00eb raport me grupin tjet\u00ebr t\u00eb t\u00eb dh\u00ebnave, shpesh sugjerohet t\u00eb p\u00ebrdoren rrjetet nervore. <\/li>\n<\/ol>\n

    <\/p>\n

    Logjika kryesore p\u00ebr k\u00ebrkimin e anomalive p\u00ebrmes t\u00eb dh\u00ebnave t\u00eb k\u00ebtyre rrjeteve nervore \u00ebsht\u00eb paraqitur n\u00eb figur\u00ebn 3:<\/p>\n

    \n\"Gjejm\u00eb
    \n<\/p>\n

    Figura 3. K\u00ebrkimi i anomalive p\u00ebrmes rrjetit nervor<\/em><\/p>\n

    Rezultati t\u00eb parashikimit ose rikuperimit t\u00eb dritares s\u00eb rrjedh\u00ebs aktuale t\u00eb metrikave llogariten devijimin nga ato t\u00eb marra nga sistemi softuerik funksional. N\u00eb rast se ka nj\u00eb ndryshim t\u00eb madh midis metrikave t\u00eb marra nga sistemi softuerik dhe rrjetit nervor, mund t\u00eb nxirret p\u00ebrfundimi p\u00ebr anomalin\u00eb e segmentit aktual t\u00eb t\u00eb dh\u00ebnave. Lind nj\u00eb s\u00ebr\u00eb problemesh p\u00ebr p\u00ebrdorimin e rrjeteve nervore:<\/p>\n

    <\/p>\n

      \n
    1. p\u00ebr t\u00eb punuar si\u00e7 duhet n\u00eb modalitetin e rrjedh\u00ebs, t\u00eb dh\u00ebnat p\u00ebr trajnimin e modeleve t\u00eb rrjeteve nervore duhet t\u00eb p\u00ebrfshijn\u00eb vet\u00ebm t\u00eb dh\u00ebna \"normale\"; <\/li>\n
    2. duhet t\u00eb kemi nj\u00eb model aktual p\u00ebr zbulimin e sakt\u00eb. Ndryshimi i trend\u00ebve dhe sezonave n\u00eb metrika mund t\u00eb shkaktoj\u00eb nj\u00eb num\u00ebr t\u00eb madh t\u00eb aktivizimeve t\u00eb gabuara t\u00eb modelit. P\u00ebr ta p\u00ebrdit\u00ebsuar at\u00eb, \u00ebsht\u00eb e nevojshme t\u00eb p\u00ebrcaktohet sakt\u00ebsisht koha kur modeli \u00ebsht\u00eb b\u00ebr\u00eb i vjetruar. N\u00ebse modeli p\u00ebrdit\u00ebsohet von\u00eb ose her\u00ebt, me siguri do t\u00eb ndodhin nj\u00eb num\u00ebr i madh i aktivizimeve t\u00eb gabuara.
      \nGjithashtu, nuk duhet harruar t\u00eb k\u00ebrkoni dhe parandaloni shfaqjen e shpesht\u00eb t\u00eb aktivizimeve t\u00eb gabuara. Supozohet se ato do t\u00eb ndodhin m\u00eb shpesh n\u00eb situata t\u00eb jasht\u00ebzakonshme. Megjithat\u00eb, ato gjithashtu mund t\u00eb jen\u00eb nj\u00eb pasoj\u00eb e gabimit t\u00eb rrjetit nervor p\u00ebr shkak t\u00eb mjaftueshm\u00ebris\u00eb s\u00eb dob\u00ebt t\u00eb trajnimit t\u00eb tij. \u00cbsht\u00eb e nevojshme t\u00eb minimizohet numri i aktivizimeve t\u00eb gabuara t\u00eb modelit. N\u00eb t\u00eb kund\u00ebrt, parashikimet e gabuara do t\u00eb shpenzojn\u00eb shum\u00eb koh\u00eb t\u00eb administratorit, e cila \u00ebsht\u00eb e destinuar p\u00ebr verifikimin e sistemit. Her\u00ebt ose von\u00eb do t\u00eb p\u00ebrfundoj\u00eb q\u00eb administratori thjesht do t\u00eb ndaloj\u00eb reagimin ndaj sistemit t\u00eb monitorimit \"paranoik\".<\/li>\n<\/ol>\n

      <\/p>\n

      Rrjeti nervor rekurent<\/h2>\n

      <\/p>\n

      P\u00ebr zbulimin e anomaliave n\u00eb serit\u00eb e koh\u00ebs mund t\u00eb aplikohet rrjeti nervor rekurent <\/a><\/noindex>me memorie LSTM. Problemi \u00ebsht\u00eb se ajo mund t\u00eb aplikohet vet\u00ebm p\u00ebr serit\u00eb e koh\u00ebs q\u00eb parashikohen. N\u00eb rastin ton\u00eb, jo t\u00eb gjitha metrikat jan\u00eb t\u00eb parashikueshme. P\u00ebrpjekja p\u00ebr t\u00eb aplikuar RNN LSTM p\u00ebr serin\u00eb e koh\u00ebs \u00ebsht\u00eb paraqitur n\u00eb figur\u00ebn 4.<\/p>\n

      \n\"Gjejm\u00eb
      \n<\/p>\n

      Figura 4. Shembulli i pun\u00ebs s\u00eb rrjetit nervor rekurent me qelizat e memories LSTM<\/em><\/p>\n

      Si\u00e7 duket nga figura 4, RNN LSTM arriti t\u00eb p\u00ebrballoj\u00eb k\u00ebrkimin e anomalive n\u00eb k\u00ebt\u00eb periudh\u00eb kohe. Aty ku rezultati ka nj\u00eb gabim t\u00eb lart\u00eb n\u00eb parashikim (gabimi mesatar), n\u00eb t\u00eb v\u00ebrtet\u00eb ka ndodhur nj\u00eb anomali n\u00eb treguesit. P\u00ebrdorimi i vet\u00ebm t\u00eb RNN LSTM \u00ebsht\u00eb qart\u00eb i pasufish\u00ebm, pasi ajo \u00ebsht\u00eb e aplikueshme p\u00ebr nj\u00eb num\u00ebr t\u00eb vog\u00ebl metrikash. Mund t\u00eb p\u00ebrdoret si nj\u00eb metod\u00eb ndihm\u00ebse p\u00ebr k\u00ebrkimin e anomalive. <\/p>\n

      <\/p>\n

      Auto-koduesi p\u00ebr parashikimin e d\u00ebshtimeve<\/h2>\n

      <\/p>\n

      Auto-koduesi<\/a><\/noindex> \u2013 n\u00eb thelb \u00ebsht\u00eb nj\u00eb rrjet nervor artificial. Shtresa hyr\u00ebse \u2013 encoder, shtresa dal\u00ebse \u2013 decoder. Disavantazhi i t\u00eb gjitha rrjeteve nervore t\u00eb k\u00ebtij lloji \u00ebsht\u00eb se ato lokalizojn\u00eb dob\u00ebt anomalit\u00eb. U zgjodh arkitektura e auto-koduesit sinkron.<\/p>\n

      \n\"Gjejm\u00eb
      \n<\/p>\n

      Figura 5. Shembulli i funksionimit t\u00eb auto-koduesit<\/em><\/p>\n

      Auto-koduesit m\u00ebsohen mbi t\u00eb dh\u00ebnat normale dhe pastaj gjejn\u00eb di\u00e7ka anormale n\u00eb t\u00eb dh\u00ebnat e paraqitura n\u00eb model. Pik\u00ebrisht ajo q\u00eb nevojitet p\u00ebr k\u00ebt\u00eb detyr\u00eb. Tani mbetet vet\u00ebm t\u00eb zgjidhet se cili nga auto-koduesit do t\u00eb p\u00ebrshtatet p\u00ebr k\u00ebt\u00eb detyr\u00eb. Forma arkitektonikisht m\u00eb e thjesht\u00eb e auto-koduesit \u00ebsht\u00eb nj\u00eb rrjet nervor i drejtp\u00ebrdrejt\u00eb, i nj\u00ebansh\u00ebm, shum\u00eb i ngjash\u00ebm me perceptronin e shum\u00ebfisht\u00eb<\/a><\/noindex> (multilayer perceptron, MLP), me nj\u00eb nivel hyr\u00ebs, nivel dal\u00ebs dhe nj\u00eb ose m\u00eb shum\u00eb shtresa t\u00eb fshehta q\u00eb i lidhin ato.
      \nMegjithat\u00eb, ndryshimet midis auto-koduesve dhe MLP q\u00ebndrojn\u00eb n\u00eb at\u00eb q\u00eb n\u00eb auto-koduesin niveli dal\u00ebs ka t\u00eb nj\u00ebjtin num\u00ebr nyjesh si niveli hyr\u00ebs, dhe se n\u00eb vend q\u00eb t\u00eb m\u00ebsohet t\u00eb parashikoj\u00eb vler\u00ebn e synuar Y, t\u00eb dh\u00ebn\u00eb nga hyrja X, auto-koduesi m\u00ebsohet t\u00eb rikonstruktoj\u00eb X-in e vet. Prandaj, auto-koduesit jan\u00eb modele t\u00eb t\u00eb m\u00ebsuarit pa mbik\u00ebqyrje. <\/p>\n

      <\/p>\n

      Detyra e auto-koduesit \u00ebsht\u00eb t\u00eb gjej\u00eb indekset e koh\u00ebs r0 \u2026 rn, q\u00eb korrespondon me elementet anormale n\u00eb vektorin hyr\u00ebs X. Ky efekt arrihet duke k\u00ebrkuar gabimin katror.<\/p>\n

      \n\"Gjejm\u00eb <\/p>\n

      Figura 6. Auto-koduesi sinkron<\/em><\/p>\n

      P\u00ebr auto-koduesin u zgjodh arkitektura sinkrone<\/a><\/noindex>. Avantazhet e saj: mund\u00ebsia e p\u00ebrdorimit t\u00eb modit t\u00eb procesimit t\u00eb rrjedh\u00ebs dhe numri relativisht m\u00eb i vog\u00ebl i parametrave t\u00eb rrjetit nervor n\u00eb krahasim me arkitektura t\u00eb tjera.<\/p>\n

      <\/p>\n

      Mekanizmi i minimizimit t\u00eb alarm\u00ebve fals<\/h2>\n

      <\/p>\n

      Duke marr\u00eb parasysh se ndodhin situata t\u00eb ndryshme jasht\u00ebzakonisht, si dhe mund\u00ebsia e munges\u00ebs s\u00eb trajnimit t\u00eb duhur t\u00eb rrjetit nervor, p\u00ebr modelin e zhvilluar t\u00eb zbulimit t\u00eb anomalive u mor vendim p\u00ebr nevoj\u00ebn e zhvillimit t\u00eb nj\u00eb mekanizmi p\u00ebr minimizimin e sinjaleve t\u00eb rreme. Ky mekaniz\u00ebm bazon n\u00eb nj\u00eb baz\u00eb t\u00eb modeleve, t\u00eb cil\u00ebn e klasifikon administratori. <\/p>\n

      <\/p>\n

      Algoritmi i transformimit dinamik t\u00eb vij\u00ebs kohore<\/a><\/noindex> (Algoritmi DTW, nga anglishtja dynamic time warping) lejon t\u00eb gjej\u00eb p\u00ebrputhjen optimale midis sekuencave t\u00eb koh\u00ebs. U aplikua p\u00ebr her\u00eb t\u00eb par\u00eb n\u00eb njohjen e z\u00ebrit: u p\u00ebrdor p\u00ebr t\u00eb p\u00ebrcaktuar se si dy sinjale t\u00eb z\u00ebrit paraqesin t\u00eb nj\u00ebjt\u00ebn fraz\u00eb t\u00eb th\u00ebn\u00eb. M\u00eb von\u00eb u gjet nj\u00eb aplikim edhe n\u00eb fusha t\u00eb tjera.<\/p>\n

      <\/p>\n

      Principi kryesor i minimizimit t\u00eb sinjaleve t\u00eb rreme \u00ebsht\u00eb mbledhja e nj\u00eb baze etalon\u00ebsh nga nj\u00eb operator, i cili klasifikon rastet e dyshuara, t\u00eb zbuluara nga rrjetet nervore. M\u00eb pas, b\u00ebhet krahasimi i etalonit t\u00eb klasifikuar me rastin q\u00eb sistemi e zbuloi, dhe merret nj\u00eb p\u00ebrfundim mbi se a i p\u00ebrket rasti nj\u00eb sinjali t\u00eb rrem\u00eb apo nj\u00eb d\u00ebshtimi. Pik\u00ebrisht p\u00ebr krahasimin e dy serive t\u00eb koh\u00ebs p\u00ebrdoret algoritmi DTW. Instrumenti kryesor p\u00ebr minimizim megjithat\u00eb mbetet klasifikimi. Supozohet se pas mbledhjes s\u00eb nj\u00eb sasie t\u00eb madhe rastesh etalon, sistemi do t\u00eb filloj\u00eb t\u00eb pyes\u00eb m\u00eb pak operatorin n\u00eb shkak t\u00eb ngjashm\u00ebris\u00eb s\u00eb shumic\u00ebs s\u00eb rasteve dhe shp\u00ebrthimit t\u00eb ngjash\u00ebm.<\/p>\n

      <\/p>\n

      Si p\u00ebrfundim, mbi baz\u00ebn e metodave t\u00eb p\u00ebrshkruara m\u00eb sip\u00ebr, u nd\u00ebrtua nj\u00eb program eksperimental p\u00ebr parashikimin e d\u00ebshtimeve t\u00eb sistemit \"Web-Konsolidimi\". Q\u00ebllimi i k\u00ebtij programi ishte, duke p\u00ebrdorur arkuimin ekzistues t\u00eb t\u00eb dh\u00ebnave t\u00eb monitorimit dhe informacionin mbi d\u00ebshtimet q\u00eb kishin ndodhur, p\u00ebr t\u00eb vler\u00ebsuar kompetenc\u00ebn e k\u00ebtij qasje p\u00ebr sistemet tona programore. Skema e pun\u00ebs s\u00eb programit p\u00ebrshkruhet m\u00eb posht\u00eb, n\u00eb figur\u00ebn 7.<\/p>\n

      \n\"Gjejm\u00eb <\/p>\n

      Figura 7. Skema e parashikimit t\u00eb d\u00ebshtimeve mbi baz\u00ebn e analiz\u00ebs s\u00eb hap\u00ebsir\u00ebs s\u00eb metrikeve<\/em><\/p>\n

      N\u00eb skem\u00eb mund t\u00eb identifikohen dy blloqe kryesore: k\u00ebrkimi i segmenteve anomale t\u00eb koh\u00ebs n\u00eb rrjedh\u00ebn e t\u00eb dh\u00ebnave t\u00eb monitorimit (metrikat) dhe mekanizmi i minimizimit t\u00eb sinjaleve t\u00eb rreme. Sh\u00ebnim: p\u00ebr q\u00ebllime eksperimentale, t\u00eb dh\u00ebnat merren p\u00ebrmes lidhjes JDBC nga baza e t\u00eb dh\u00ebnave, n\u00eb t\u00eb cil\u00ebn ruhet graphite.
      \nK\u00ebtu \u00ebsht\u00eb nd\u00ebrfaqja e sistemit t\u00eb monitorimit t\u00eb zhvilluar (figura 8).<\/p>\n

      \n\"Gjejm\u00eb<\/p>\n

      Figura 8. Nd\u00ebrfaqja e sistemit eksperimental t\u00eb monitorimit<\/em><\/p>\n

      N\u00eb nd\u00ebrfaqe tregohet p\u00ebrqindja e anomalis\u00eb p\u00ebr metrikat e marra. N\u00eb rastin ton\u00eb, marrja simulohet. Ne tashm\u00eb kemi t\u00eb gjitha t\u00eb dh\u00ebnat p\u00ebr disa jav\u00eb dhe po i ngarkojm\u00eb gradualisht p\u00ebr t\u00eb verifikuar rastin me anomali q\u00eb \u00e7on n\u00eb d\u00ebshtim. N\u00eb shiritin e statusit t\u00eb posht\u00ebm tregohet p\u00ebrqindja totale e anomalis\u00eb s\u00eb t\u00eb dh\u00ebnave n\u00eb k\u00ebt\u00eb moment, e cila p\u00ebrcaktohet me ndihm\u00ebn e auto-koduesit. Po ashtu, p\u00ebr metrikat e parashikuara tregohet nj\u00eb p\u00ebrqindje e ve\u00e7ant\u00eb, e cila llogaritet nga RNN LSTM.<\/p>\n

      <\/p>\n

      Shembulli i zbulimit t\u00eb anomalis\u00eb n\u00eb treguesit e CPU-s\u00eb me ndihm\u00ebn e rrjetit nervor RNN LSTM (figura 9).<\/p>\n

      \n\"Gjejm\u00eb<\/p>\n

      Figura 9. Zbulimi me RNN LSTM<\/em><\/p>\n

      Nj\u00eb rast mjaft i thjesht\u00eb, n\u00eb thelb nj\u00eb shp\u00ebrthim i zakonsh\u00ebm, por q\u00eb \u00e7on n\u00eb d\u00ebshtimin e sistemit, u zbulua me sukses me ndihm\u00ebn e RNN LSTM. Treguesi i anomalis\u00eb n\u00eb k\u00ebt\u00eb periudh\u00eb kohore \u00ebsht\u00eb 85 \u2013 95%, gjith\u00e7ka mbi 80% (kjo prag \u00ebsht\u00eb p\u00ebrcaktuar eksperimentalisht) konsiderohet nj\u00eb anomali.
      \nShembulli i zbulimit t\u00eb anomalis\u00eb kur sistemi nuk ishte n\u00eb gjendje t\u00eb ngarkohej pas p\u00ebrdit\u00ebsimit. Kjo situat\u00eb identifikohet nga auto-koduesi (figura 10).<\/p>\n

      \n\"Gjejm\u00eb<\/p>\n

      Figura 10. Shembulli i zbulimit nga auto-koduesi<\/em><\/p>\n

      Si\u00e7 duket nga figura, PermGen u ngat\u00ebrrua n\u00eb nj\u00eb nivel. Auto-koduesi e konsideroi k\u00ebt\u00eb t\u00eb \u00e7uditsh\u00ebm, pasi nuk kishte par\u00eb asgj\u00eb t\u00eb ngjashme m\u00eb par\u00eb. K\u00ebtu anomalia mbahet n\u00eb 100% deri n\u00eb rikthimin e sistemit n\u00eb gjendje funksionale. Anomalia tregohet p\u00ebr t\u00eb gjitha metrikat. Si\u00e7 u p\u00ebrmend m\u00eb par\u00eb, auto-koduesi nuk \u00ebsht\u00eb n\u00eb gjendje t\u00eb lokalizoj\u00eb anomali. Operator\u00ebt jan\u00eb t\u00eb thirrur p\u00ebr t\u00eb kryer k\u00ebt\u00eb funksion n\u00eb k\u00ebto situata.<\/p>\n

      <\/p>\n

      P\u00ebrfundim<\/h2>\n

      <\/p>\n

      PC \u00abWeb-Konsolidimi\u00bb po zhvillohet p\u00ebr disa vite. Sistemi \u00ebsht\u00eb n\u00eb nj\u00eb gjendje mjaft t\u00eb q\u00ebndrueshme dhe numri i incidenteve t\u00eb regjistruara \u00ebsht\u00eb i vog\u00ebl. Megjithat\u00eb, \u00ebsht\u00eb arritur t\u00eb gjenden anomali q\u00eb \u00e7ojn\u00eb n\u00eb d\u00ebshtim p\u00ebr 5 \u2013 10 minuta para se ndodhi d\u00ebshtimi. N\u00eb disa raste, njoftimi p\u00ebr d\u00ebshtim do t\u00eb ndihmonte t\u00eb kurseni koh\u00ebn e parapar\u00eb q\u00eb i \u00ebsht\u00eb kushtuar kryerjes s\u00eb punimeve \"reparimtare\".<\/p>\n

      <\/p>\n

      N\u00eb eksperimet q\u00eb arrit\u00ebm t\u00eb kryejm\u00eb, \u00ebsht\u00eb akoma her\u00ebt p\u00ebr t\u00eb b\u00ebr\u00eb p\u00ebrfundime p\u00ebrfundimtare. Deri m\u00eb tani, rezultatet jan\u00eb kontradiktore. Nga nj\u00ebra an\u00eb, \u00ebsht\u00eb e qart\u00eb se algoritmet e bazuara n\u00eb rrjete neurale jan\u00eb n\u00eb gjendje t\u00eb zbulojn\u00eb anomali \"t\u00eb dobishme\". Nga ana tjet\u00ebr, mbetet nj\u00eb p\u00ebrqindje e madhe e alarm\u00ebve t\u00eb rrem\u00eb, dhe nuk t\u00eb gjitha anomali q\u00eb nj\u00eb specialist i kualifikuar i zbulohet rrjetit nervor arrijn\u00eb t\u00eb duken. Nj\u00eb mang\u00ebsi \u00ebsht\u00eb se tani p\u00ebr tani rrjeti nervor k\u00ebrkon m\u00ebsim t\u00eb udh\u00ebhequr p\u00ebr t\u00eb funksionuar normalisht.<\/p>\n

      <\/p>\n

      P\u00ebr zhvillimin e m\u00ebtejsh\u00ebm t\u00eb sistemit t\u00eb parashikimit t\u00eb defekteve dhe p\u00ebr t'i dh\u00ebn\u00eb atij nj\u00eb gjendje t\u00eb k\u00ebnaqshme, mund t\u00eb parashikohet disa rrug\u00eb. Kjo \u00ebsht\u00eb nj\u00eb analiz\u00eb m\u00eb e detajuar e rasteve me anomali q\u00eb \u00e7ojn\u00eb n\u00eb d\u00ebshtim, duke e pasuruar k\u00ebshtu list\u00ebn e metrikave t\u00eb r\u00ebnd\u00ebsishme q\u00eb ndikojn\u00eb shum\u00eb n\u00eb gjendjen e sistemit, dhe duke hequr ato t\u00eb tep\u00ebrta q\u00eb nuk kan\u00eb ndikim n\u00eb t\u00eb. Gjithashtu, n\u00ebse ecet n\u00eb k\u00ebt\u00eb drejtim, mund t\u00eb b\u00ebhen p\u00ebrpjekje p\u00ebr t\u00eb specializuar algoritmet konkretisht p\u00ebr rastet tona me anomali q\u00eb \u00e7ojn\u00eb n\u00eb defekte. Ka edhe nj\u00eb rrug\u00eb tjet\u00ebr. Kjo \u00ebsht\u00eb p\u00ebrmir\u00ebsimi i strukturave t\u00eb rrjeteve neurale dhe rritja e sakt\u00ebsis\u00eb s\u00eb zbulesave duke reduktuar koh\u00ebn e m\u00ebsimit.<\/p>\n

      <\/p>\n

      Shpreh mir\u00ebnjohjen time ndaj koleg\u00ebve q\u00eb m\u00eb ndihmuan me shkruan dhe mbajtjen e aktualitetit t\u00eb k\u00ebtij artikulli: Viktor Verbitsky<\/a><\/noindex> dhe Sergey Finogenov.<\/p>\n

      Burimi: habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

      \u041f\u0440\u043e\u043c\u044b\u0448\u043b\u0435\u043d\u043d\u0430\u044f \u0440\u0430\u0437\u0440\u0430\u0431\u043e\u0442\u043a\u0430 \u043f\u0440\u043e\u0433\u0440\u0430\u043c\u043c\u043d\u044b\u0445 \u0441\u0438\u0441\u0442\u0435\u043c \u0442\u0440\u0435\u0431\u0443\u0435\u0442 \u0431\u043e\u043b\u044c\u0448\u043e\u0433\u043e \u0432\u043d\u0438\u043c\u0430\u043d\u0438\u044f \u043a \u043e\u0442\u043a\u0430\u0437\u043e\u0443\u0441\u0442\u043e\u0439\u0447\u0438\u0432\u043e\u0441\u0442\u0438 \u043a\u043e\u043d\u0435\u0447\u043d\u043e\u0433\u043e \u043f\u0440\u043e\u0434\u0443\u043a\u0442\u0430, \u0430 \u0442\u0430\u043a\u0436\u0435 \u0431\u044b\u0441\u0442\u0440\u043e\u0433\u043e \u0440\u0435\u0430\u0433\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f \u043d\u0430 \u043e\u0442\u043a\u0430\u0437\u044b \u0438 \u0441\u0431\u043e\u0438, \u0435\u0441\u043b\u0438 \u043e\u043d\u0438 \u0432\u0441\u0435-\u0442\u0430\u043a\u0438 \u0441\u043b\u0443\u0447\u0430\u044e\u0442\u0441\u044f. \u041c\u043e\u043d\u0438\u0442\u043e\u0440\u0438\u043d\u0433, \u043a\u043e\u043d\u0435\u0447\u043d\u043e \u0436\u0435, \u043f\u043e\u043c\u043e\u0433\u0430\u0435\u0442 \u0440\u0435\u0430\u0433\u0438\u0440\u043e\u0432\u0430\u0442\u044c \u043d\u0430 \u043e\u0442\u043a\u0430\u0437\u044b \u0438 \u0441\u0431\u043e\u0438 \u044d\u0444\u0444\u0435\u043a\u0442\u0438\u0432\u043d\u0435\u0435 \u0438 \u0431\u044b\u0441\u0442\u0440\u0435\u0435, \u043d\u043e \u043d\u0435\u0434\u043e\u0441\u0442\u0430\u0442\u043e\u0447\u043d\u043e. \u0412\u043e-\u043f\u0435\u0440\u0432\u044b\u0445, \u043e\u0447\u0435\u043d\u044c \u0441\u043b\u043e\u0436\u043d\u043e \u0443\u0441\u043b\u0435\u0434\u0438\u0442\u044c \u0437\u0430 \u0431\u043e\u043b\u044c\u0448\u0438\u043c \u043a\u043e\u043b\u0438\u0447\u0435\u0441\u0442\u0432\u043e\u043c \u0441\u0435\u0440\u0432\u0435\u0440\u043e\u0432 \u2013 \u043d\u0435\u043e\u0431\u0445\u043e\u0434\u0438\u043c\u043e \u0431\u043e\u043b\u044c\u0448\u043e\u0435 \u043a\u043e\u043b\u0438\u0447\u0435\u0441\u0442\u0432\u043e \u043b\u044e\u0434\u0435\u0439. \u0412\u043e-\u0432\u0442\u043e\u0440\u044b\u0445, \u043d\u0443\u0436\u043d\u043e \u0445\u043e\u0440\u043e\u0448\u043e \u043f\u043e\u043d\u0438\u043c\u0430\u0442\u044c, \u043a\u0430\u043a […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[688],"tags":[],"class_list":["post-53531","post","type-post","status-publish","format-standard","hentry","category-administrirovanie"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47K\u00ebrkojm\u00eb anomali dhe parashikojm\u00eb d\u00ebshtime me ndihm\u00ebn e rrjeteve neurale | ProHoster","description":"Zhvillimi industrial.","canonical_url":"https:\/\/prohoster.info\/sq\/blog\/administrirovanie\/ishhem-anomalii-i-predskazyvaem-sboi-s-pomoshhyu-nejrosetej","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"sq_AL","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u0418\u0449\u0435\u043c \u0430\u043d\u043e\u043c\u0430\u043b\u0438\u0438 \u0438 \u043f\u0440\u0435\u0434\u0441\u043a\u0430\u0437\u044b\u0432\u0430\u0435\u043c \u0441\u0431\u043e\u0438 \u0441 \u043f\u043e\u043c\u043e\u0449\u044c\u044e \u043d\u0435\u0439\u0440\u043e\u0441\u0435\u0442\u0435\u0439 | ProHoster","og:description":"\u041f\u0440\u043e\u043c\u044b\u0448\u043b\u0435\u043d\u043d\u0430\u044f \u0440\u0430\u0437\u0440\u0430\u0431\u043e\u0442\u043a\u0430.","og:url":"https:\/\/prohoster.info\/sq\/blog\/administrirovanie\/ishhem-anomalii-i-predskazyvaem-sboi-s-pomoshhyu-nejrosetej","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2019-12-03T21:00:00+00:00","article:modified_time":"2020-02-18T11:01:26+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"53531","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":"2026-01-24 07:46:24","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-02-28 20:23:55","updated":"2026-01-24 07:46:24","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts\/53531","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/comments?post=53531"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts\/53531\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/media?parent=53531"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/categories?post=53531"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/tags?post=53531"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}