Modèle de Survival SVM - Prédictions conformes

Imports¶

import polars as pl

Chargement des données¶

from sksurv.util import Surv

df_train = pl.read_parquet("../../data/df_study_L18_w6_train.parquet")
df_test = pl.read_parquet("../../data/df_study_L18_w6_test.parquet")

X_train = df_train.drop(["event", "time"]).to_pandas()
X_test = df_test.drop(["event", "time"]).to_pandas()

y_train = Surv.from_dataframe("event", "time", df_train.to_pandas())
y_test = Surv.from_dataframe("event", "time", df_test.to_pandas())

Entraînement du modèle¶

from sksurv.svm import FastSurvivalSVM

svm = FastSurvivalSVM(rank_ratio=1.0)
svm.fit(X_train, y_train);

Évaluation du modèle final¶

from utils import evaluate_survival_model

risk_svm = svm.predict(X_test)

evaluate_survival_model(df_train, df_test, risk_svm)