- \n
- Duyarl\u0131l\u0131k:<\/strong> Sonu\u00e7 ger\u00e7ekten olumluyken modelin bir g\u00f6zlem i\u00e7in olumlu bir sonu\u00e7 tahmin etme olas\u0131l\u0131\u011f\u0131. Buna ayn\u0131 zamanda \u201cger\u00e7ek pozitif oran\u201d da denir.<\/span><\/li>\n
- \u00d6zg\u00fcll\u00fck:<\/strong> Sonu\u00e7 ger\u00e7ekten negatif oldu\u011funda modelin bir g\u00f6zlem i\u00e7in negatif bir sonu\u00e7 \u00f6ng\u00f6rme olas\u0131l\u0131\u011f\u0131. Buna ayn\u0131 zamanda \u201cger\u00e7ek negatif oran\u201d da denir.<\/span><\/li>\n<\/ul>\n
Bu iki \u00f6l\u00e7\u00fcm\u00fc g\u00f6rselle\u015ftirmenin bir yolu, “al\u0131c\u0131 \u00e7al\u0131\u015fma karakteristi\u011fi” e\u011frisi anlam\u0131na gelen bir ROC e\u011frisi<\/strong> olu\u015fturmakt\u0131r. Bu, lojistik regresyon modelinin duyarl\u0131l\u0131\u011f\u0131n\u0131 ve \u00f6zg\u00fcll\u00fc\u011f\u00fcn\u00fc g\u00f6steren bir grafiktir.<\/span><\/p>\n
A\u015fa\u011f\u0131daki ad\u0131m ad\u0131m \u00f6rnek, Python’da bir ROC e\u011frisinin nas\u0131l olu\u015fturulaca\u011f\u0131n\u0131 ve yorumlanaca\u011f\u0131n\u0131 g\u00f6sterir.<\/span><\/p>\n
Ad\u0131m 1: Gerekli paketleri i\u00e7e aktar\u0131n<\/span><\/strong><\/span><\/h3>\n
\u00d6ncelikle Python’da lojistik regresyon ger\u00e7ekle\u015ftirmek i\u00e7in gerekli paketleri i\u00e7e aktaraca\u011f\u0131z:<\/span><\/p>\n
import<\/span> pandas as<\/span> pd\nimport<\/span> numpy as<\/span> np\nfrom<\/span> sklearn. model_selection<\/span> import<\/span> train_test_split\nfrom<\/span> sklearn. linear_model<\/span> import<\/span> LogisticRegression\nfrom<\/span> sklearn import<\/span> metrics\nimport<\/span> matplotlib. pyplot<\/span> as<\/span> plt\n<\/strong><\/pre>\nAd\u0131m 2: Lojistik regresyon modelini yerle\u015ftirin<\/strong><\/span><\/h3>\n
Daha sonra, bir veri k\u00fcmesini i\u00e7e aktaraca\u011f\u0131z ve buna bir lojistik regresyon modeli yerle\u015ftirece\u011fiz:<\/span><\/p>\n
#import dataset from CSV file on Github\n<\/span>url = \"https:\/\/raw.githubusercontent.com\/Statorials\/Python-Guides\/main\/default.csv\"\ndata = pd. read_csv<\/span> (url)\n\n#define the predictor variables and the response variable\n<\/span>X = data[[' student<\/span> ',' balance<\/span> ',' income<\/span> ']]\ny = data[' default<\/span> ']\n\n#split the dataset into training (70%) and testing (30%) sets\n<\/span>X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.3,random_state=0) \n\n#instantiate the model\n<\/span>log_regression = LogisticRegression()\n\n#fit the model using the training data\n<\/span>log_regression. fit<\/span> (X_train,y_train)<\/strong><\/span><\/pre>\nAd\u0131m 3: ROC e\u011frisini \u00e7izin<\/strong><\/span><\/h3>\n
Daha sonra, ger\u00e7ek pozitif oran\u0131 ve yanl\u0131\u015f pozitif oran\u0131 hesaplayaca\u011f\u0131z ve Matplotlib veri g\u00f6rselle\u015ftirme paketini kullanarak bir ROC e\u011frisi olu\u015fturaca\u011f\u0131z:<\/span> <\/p>\n
#define metrics\n<\/span>y_pred_proba = log_regression. predict_proba<\/span> (X_test)[::,1]\nfpr, tpr, _ = metrics. roc_curve<\/span> (y_test, y_pred_proba)\n\n#create ROC curve\n<\/span>plt. plot<\/span> (fpr,tpr)\nplt. ylabel<\/span> (' True Positive Rate<\/span> ')\nplt. xlabel<\/span> (' False Positive Rate<\/span> ')\nplt. show<\/span> ()<\/strong><\/span> <\/pre>\n![\"\"](\"https:\/\/statorials.org\/wp-content\/uploads\/2023\/08\/rocpython1.png\")
<\/p>\nE\u011fri grafi\u011fin sol \u00fcst k\u00f6\u015fesine ne kadar yak\u0131n olursa, model verileri kategorilere g\u00f6re o kadar iyi s\u0131n\u0131fland\u0131rabilir.<\/span><\/p>\n
Yukar\u0131daki grafikten g\u00f6rebilece\u011fimiz gibi, bu lojistik regresyon modeli, verileri kategorilere ay\u0131rma konusunda olduk\u00e7a zay\u0131f bir i\u015f \u00e7\u0131kar\u0131yor.<\/span><\/p>\n
Bunu \u00f6l\u00e7mek i\u00e7in, bize grafi\u011fin ne kadar\u0131n\u0131n e\u011frinin alt\u0131nda oldu\u011funu s\u00f6yleyen AUC’yi (e\u011frinin alt\u0131ndaki alan\u0131) hesaplayabiliriz.<\/span><\/p>\n
AUC 1’e ne kadar yak\u0131nsa model o kadar iyidir. AUC de\u011feri 0,5’e e\u015fit olan bir model, rastgele s\u0131n\u0131fland\u0131rma yapan bir modelden daha iyi de\u011fildir.<\/span><\/p>\n
Ad\u0131m 4: AUC’yi hesaplay\u0131n<\/strong><\/span><\/h3>\n
- \u00d6zg\u00fcll\u00fck:<\/strong> Sonu\u00e7 ger\u00e7ekten negatif oldu\u011funda modelin bir g\u00f6zlem i\u00e7in negatif bir sonu\u00e7 \u00f6ng\u00f6rme olas\u0131l\u0131\u011f\u0131. Buna ayn\u0131 zamanda \u201cger\u00e7ek negatif oran\u201d da denir.<\/span><\/li>\n<\/ul>\n
![\"\"](\"https:\/\/statorials.org\/wp-content\/uploads\/2023\/08\/rocpython2.png\")
<\/p>\n