def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
params = clf.get_params()
logging.info("Selected XGBoost Params are %s " % params)
safe_del(ret, "xgboost")
ret["xgboost"] = {}
ret["xgboost"]["max_depth"] = params["max_depth"]
ret["xgboost"]["learning_rate"] = params["learning_rate"]
ret["xgboost"]["n_estimators"] = params["n_estimators"]
ret["xgboost"]["nthread"] = params["n_jobs"] if params[
"n_jobs"] > 0 else -1 # TODO: change => migration ?
ret["xgboost"]["gamma"] = params["gamma"]
ret["xgboost"]["min_child_weight"] = params["min_child_weight"]
ret["xgboost"]["max_delta_step"] = params["max_delta_step"]
ret["xgboost"]["subsample"] = params["subsample"]
ret["xgboost"]["colsample_bytree"] = params["colsample_bytree"]
ret["xgboost"]["colsample_bylevel"] = params["colsample_bylevel"]
ret["xgboost"]["alpha"] = params["reg_alpha"]
ret["xgboost"]["lambda"] = params["reg_lambda"]
ret["xgboost"]["seed"] = params[
"random_state"] # TODO: change => migration ?
ret["xgboost"]["impute_missing"] = True if params["missing"] else False
ret["xgboost"]["missing"] = params["missing"]
ret["xgboost"]["base_score"] = params["base_score"]
ret["xgboost"]["scale_pos_weight"] = params["scale_pos_weight"]
ret["xgboost"]["enable_early_stopping"] = fit_params.get(
'early_stopping_rounds') is not None
ret["xgboost"]["early_stopping_rounds"] = fit_params.get(
'early_stopping_rounds')
ret["xgboost"]["booster"] = params.get("booster")
ret["xgboost"]["objective"] = params.get("objective").replace(":", "_")
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
params = clf.get_params()
logging.info("Selected Ordinary Least Squares Params are %s " % params)
safe_del(ret, "least_squares")
ret["n_jobs"] = params["n_jobs"]
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
safe_del(ret, "ridge_grid")
params = clf.get_params()
ret["lasso"] = {}
if hasattr(clf, "alpha_"):
ret["lasso"]["alpha"] = params.get("alpha", clf.alpha_)
else:
ret["lasso"]["alpha"] = params.get("alpha", 0)
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
safe_del(ret, "logit_grid")
params = clf.get_params()
logging.info("LR Params are %s " % params)
ret["logit"] = {
"penalty": params["penalty"],
"multi_class": params["multi_class"],
"C": params["C"]
}
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
safe_del(ret, "dtc_classifier_grid")
params = clf.get_params()
logging.info("DT params are %s " % params)
ret["dt"] = {
"max_depth": params["max_depth"],
"criterion": params["criterion"],
"min_samples_leaf": params["min_samples_leaf"],
"splitter": params["splitter"]
}
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
params = clf.get_params()
logging.info("Selected KNN Params are %s " % params)
safe_del(ret, "knn_grid")
ret["knn"] = {
"k": params["n_neighbors"],
"distance_weighting": params["weights"] == "distance",
"algorithm": params["algorithm"],
"p": params["p"],
"leaf_size": params["leaf_size"],
}
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
params = clf.get_params()
logging.info("Selected SGD Params are %s " % params)
safe_del(ret, "sgd_grid")
ret["sgd"] = {
"loss": params["loss"],
"penalty": params["penalty"],
"alpha": params["alpha"],
"l1_ratio": params["l1_ratio"],
"n_jobs": params["n_jobs"],
"n_iter": clf.n_iter_
}
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
params = clf.get_params()
logging.info("Selected SVC Params are %s " % params)
safe_del(ret, "svc_grid")
ret["svm"] = {
"C": params["C"],
"kernel": params["kernel"],
"gamma": 0.0 if params["gamma"] == 'auto' else params["gamma"],
"tol": params["tol"],
"max_iter": params["max_iter"],
"coef0": params["coef0"]
}
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
safe_del(ret, "extra_trees_grid")
params = clf.get_params()
logging.info("Extra trees Params are %s " % params)
ret["extra_trees"] = {
"estimators": len(clf.estimators_),
"njobs": params["n_jobs"] if params["n_jobs"] > 0 else -1,
"max_tree_depth": params["max_depth"],
"min_samples_leaf": params["min_samples_leaf"],
"selection_mode": get_selection_mode(params["max_features"]),
}
if ret["extra_trees"]["selection_mode"] == "number":
ret["extra_trees"]["max_features"] = clf.max_features
if ret["extra_trees"]["selection_mode"] == "prop":
ret["extra_trees"]["max_feature_prop"] = clf.max_features
amp["other"]["rf_min_samples_split"] = params["min_samples_split"]
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
safe_del(ret, "gbt_classifier_grid")
params = clf.get_params()
logging.info("GBT Params are %s " % params)
ret["gbt"] = {
"n_estimators": len(clf.estimators_),
"max_depth": params["max_depth"],
"learning_rate": params["learning_rate"],
"min_samples_leaf": params["min_samples_leaf"],
"selection_mode": get_selection_mode(params["max_features"]),
"loss": params["loss"]
}
if ret["gbt"]["selection_mode"] == "number":
ret["gbt"]["max_features"] = ret["gbt_selection_mode"]
if ret["gbt"]["selection_mode"] == "prop":
ret["gbt"]["max_feature_prop"] = ret["gbt_selection_mode"]
return amp
def output_params(self, ret, clf, fit_params):
amp = {"resolved": ret, "other": {}}
safe_del(ret, "rf_classifier_grid")
params = clf.get_params()
logging.info("Obtained RF CLF params: %s " % params)
ret["rf"] = {
"estimators": len(clf.estimators_),
"max_tree_depth": params["max_depth"],
"min_samples_leaf": params["min_samples_leaf"],
"selection_mode": get_selection_mode(params["max_features"]),
}
if ret["rf"]["selection_mode"] == "number":
ret["rf"]["max_features"] = clf.max_features
if ret["rf"]["selection_mode"] == "prop":
ret["rf"]["max_feature_prop"] = clf.max_features
amp["other"]["rf_min_samples_split"] = params["min_samples_split"]
return amp