def gen_X(self, X, lags, start):
factors = np.asanyarray(self.factors.loc[X.index])
# remove factors that have nulls values
# factors = factors.ix[:, factors.isnull().sum(axis=0)==0]
# print(factors[:,~np.all(np.isnan(factors), axis=0)])
factors = factors[:, ~np.any(np.isnan(factors), axis=0)]
# print(np.isnan(factors))
# center and standardise
f_mean = factors.mean()
f_std = factors.std()
factors = (factors - f_mean) / f_std
if start == 0: # train
# fit PCA projection
self.pca = PCA(n_components=self.n_pc)
self.pca.fit(factors)
# else, test: reuse factor projection
f = pd.DataFrame(self.pca.transform(factors), index=X.index)
# add factors after the usual columns
X = pd.concat([X, f], axis=1)
return VAR.gen_X(X, lags, start)
def gen_y(y, start):
y = VAR.gen_y(y, lags, start)
return y
def gen_x(X, start):
X = VAR.gen_X(X, lags, start)
# X = X[:, 1:] # remove constant
return X
def gen_y(self, y, lags, start):
return VAR.gen_y(y, lags, start)