Her iki \u00f6l\u00e7\u00fcm i\u00e7in de de\u011fer ne kadar y\u00fcksek olursa, model g\u00f6zlemleri s\u0131n\u0131flara ay\u0131rma konusunda o kadar yetenekli olur.<\/span><\/p>\n Ancak her metrik farkl\u0131 bir form\u00fcl kullan\u0131larak hesaplan\u0131r ve onu kullanman\u0131n avantaj ve dezavantajlar\u0131 vard\u0131r.<\/span><\/p>\n A\u015fa\u011f\u0131daki \u00f6rnek, her bir metri\u011fin pratikte nas\u0131l hesaplanaca\u011f\u0131n\u0131 g\u00f6stermektedir.<\/span><\/p>\n 400 farkl\u0131 kolej basketbol oyuncusunun NBA’e al\u0131n\u0131p al\u0131nmayaca\u011f\u0131n\u0131 tahmin etmek i\u00e7in lojistik regresyon modeli kulland\u0131\u011f\u0131m\u0131z\u0131 varsayal\u0131m.<\/span><\/p>\n A\u015fa\u011f\u0131daki kar\u0131\u015f\u0131kl\u0131k matrisi, modelin yapt\u0131\u011f\u0131 tahminleri \u00f6zetlemektedir:<\/span> <\/p>\n Kar\u0131\u015f\u0131kl\u0131k matrisi i\u00e7in \u00e7e\u015fitli \u00f6l\u00e7\u00fcmleri nas\u0131l hesaplayaca\u011f\u0131n\u0131z a\u015fa\u011f\u0131da a\u00e7\u0131klanm\u0131\u015ft\u0131r:<\/span><\/p>\n Do\u011fruluk:<\/strong> pozitif tahminlerin toplam pozitif tahminlere g\u00f6re do\u011fru olmas\u0131<\/span><\/p>\n Hat\u0131rlatma:<\/strong> Pozitif tahminleri toplam ger\u00e7ek pozitiflere g\u00f6re d\u00fczeltin<\/span><\/p>\n Do\u011fruluk:<\/strong> Do\u011fru \u015fekilde s\u0131n\u0131fland\u0131r\u0131lan t\u00fcm g\u00f6zlemlerin y\u00fczdesi<\/span><\/p>\n F1 puan\u0131:<\/strong> hassasl\u0131k ve geri \u00e7a\u011f\u0131rman\u0131n harmonik ortalamas\u0131<\/span><\/p>\n F1 skorunu ve do\u011frulu\u011funu kullanman\u0131n art\u0131lar\u0131 ve eksileri var.<\/span><\/p>\n Do\u011fruluk<\/strong> :<\/span><\/p>\n Art\u0131lar\u0131<\/strong> : Yorumlanmas\u0131 kolayd\u0131r. Bir modelin %90 do\u011fru oldu\u011funu s\u00f6ylersek, g\u00f6zlemlerin %90’\u0131n\u0131 do\u011fru s\u0131n\u0131fland\u0131rd\u0131\u011f\u0131n\u0131 biliyoruz.<\/span><\/p>\n Dezavantaj\u0131<\/strong> : Verilerin nas\u0131l da\u011f\u0131t\u0131ld\u0131\u011f\u0131n\u0131 dikkate almaz. \u00d6rne\u011fin, oyuncular\u0131n %90’\u0131n\u0131n NBA’e se\u00e7ilmedi\u011fini varsayal\u0131m. Basit\u00e7e her oyuncunun draft edilmeyece\u011fini \u00f6ng\u00f6ren bir modelimiz olsayd\u0131, model oyuncular\u0131n %90’\u0131n\u0131n sonucunu do\u011fru bir \u015fekilde tahmin ederdi. Bu de\u011fer y\u00fcksek gibi g\u00f6r\u00fcnse de asl\u0131nda model hangi oyuncular\u0131n draft edilece\u011fini do\u011fru bir \u015fekilde tahmin edemiyor.<\/span><\/p>\n F1 sonu\u00e7lar\u0131<\/strong> :<\/span><\/p>\n Art\u0131lar\u0131<\/strong> : Verilerin nas\u0131l da\u011f\u0131t\u0131ld\u0131\u011f\u0131n\u0131 d\u00fc\u015f\u00fcn\u00fcn. \u00d6rne\u011fin, veriler olduk\u00e7a dengesizse (\u00f6rne\u011fin t\u00fcm oyuncular\u0131n %90’\u0131 se\u00e7ilmemi\u015f ve %10’u se\u00e7ilmi\u015fse), o zaman F1 puan\u0131 modelin performans\u0131n\u0131n daha iyi bir \u015fekilde de\u011ferlendirilmesini sa\u011flayacakt\u0131r.<\/span><\/p>\n Dezavantaj\u0131<\/strong> : Yorumlanmas\u0131 daha zordur. F1 puan\u0131, hassasiyet ve model hat\u0131rlaman\u0131n bir kar\u0131\u015f\u0131m\u0131d\u0131r ve yorumlanmas\u0131n\u0131 biraz daha zorla\u015ft\u0131r\u0131r.<\/span><\/p>\n Genel olarak:<\/span><\/p>\n S\u0131n\u0131flar dengeli oldu\u011funda ve yanl\u0131\u015f negatifleri tahmin etmenin b\u00fcy\u00fck bir dezavantaj\u0131 olmad\u0131\u011f\u0131nda do\u011frulu\u011fu<\/strong> s\u0131kl\u0131kla kullan\u0131r\u0131z.<\/span><\/p>\n S\u0131n\u0131flar dengesiz oldu\u011funda ve yanl\u0131\u015f negatifleri tahmin etmede ciddi bir dezavantaj oldu\u011funda F1 puan\u0131n\u0131<\/strong> s\u0131kl\u0131kla kullan\u0131r\u0131z.<\/span><\/p>\n \u00d6rne\u011fin, bir ki\u015finin kanser olup olmad\u0131\u011f\u0131n\u0131 tahmin etmek i\u00e7in bir lojistik regresyon modeli kullan\u0131rsak, yanl\u0131\u015f negatifler ger\u00e7ekten k\u00f6t\u00fcd\u00fcr (\u00f6rne\u011fin, bir ki\u015finin ger\u00e7ekten kanser hastas\u0131 oldu\u011funu tahmin etmek a), bu nedenle F1 puan\u0131, kansere sahip olan modelleri cezaland\u0131racakt\u0131r. \u00e7ok fazla yanl\u0131\u015f negatif. kesinlikten daha fazlas\u0131.<\/span><\/p>\n Regresyon vs. s\u0131n\u0131fland\u0131rma: fark nedir?<\/a> Makine \u00f6\u011freniminde s\u0131n\u0131fland\u0131rma modellerini kullan\u0131rken model kalitesini de\u011ferlendirmek i\u00e7in s\u0131kl\u0131kla kulland\u0131\u011f\u0131m\u0131z iki \u00f6l\u00e7\u00fcm F1 puan\u0131 ve do\u011fruluktur . Her iki \u00f6l\u00e7\u00fcm i\u00e7in de de\u011fer ne kadar y\u00fcksek olursa, model g\u00f6zlemleri s\u0131n\u0131flara ay\u0131rma konusunda o kadar yetenekli olur. Ancak her metrik farkl\u0131 bir form\u00fcl kullan\u0131larak hesaplan\u0131r ve onu kullanman\u0131n avantaj ve dezavantajlar\u0131 vard\u0131r. A\u015fa\u011f\u0131daki \u00f6rnek, her […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[],"class_list":["post-2164","post","type-post","status-publish","format-standard","hentry","category-rehber"],"yoast_head":"\n \u00d6rnek: F1 puan\u0131n\u0131n ve do\u011frulu\u011funun hesaplanmas\u0131<\/strong><\/span><\/h3>\n
![\"\"](\"https:\/\/statorials.org\/wp-content\/uploads\/2023\/08\/\u1405-\u1405-\u1405.png\")
<\/p>\n\n
\n
\n
\n
Do\u011frulu\u011fa kar\u015f\u0131 F1 puan\u0131 ne zaman kullan\u0131lmal\u0131?<\/strong><\/span><\/h3>\n
Ek kaynaklar<\/strong><\/span><\/h3>\n
Lojistik Regresyona Giri\u015f<\/a>
R’de lojistik regresyon nas\u0131l ger\u00e7ekle\u015ftirilir
Python’da Lojistik Regresyon Nas\u0131l Ger\u00e7ekle\u015ftirilir<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"