class KernelModel:
def __init__(self, dh=None, kernel_cols=None):
if dh is None:
self.dh = DataHandler()
else:
self.dh = dh
if kernel_cols is None:
self.kernel_cols = "all"
else:
self.kernel_cols = kernel_cols
logging.info("Initializing")
self.kde = None
def train_model(self):
if self.kernel_cols != "all":
vals = self.dh.full_df[self.kernel_cols].values.T
else:
vals = self.dh.get_density_df().values.T
self.kde = stats.gaussian_kde(vals)
logging.info("Trained a kde")
"""
Assumes a preprocessed df sent
in and predicts, for each data point,
a density prediction.
"""
def get_densities(self, df=None):
if self.kde is None:
logging.info("No trained kde, training now. ")
self.train_model()
if df is None:
df = self.dh.full_df
if self.kernel_cols != "all":
vals = df[self.kernel_cols].values.T
else:
vals = df.values.T
return self.kde.evaluate(vals)
