基于LSTM神经网络的风力发电风速预测 python源码+项目说明.zip

#模糊粗糙集的属性约简 #参考github和matlab模糊粗糙集算法 import numpy as np class RoughSetsReducer: def __size(self, x): return (1, x.shape[0]) if x.ndim == 1 else x.shape ''' Calculates indiscernibility relation ''' def indisc(self, a, x): def codea(a, x, b): yy = 0 for i in range(0, a): yy += x[i] * b**(a-(i+1)) return yy p, q = self.__size(x) ap, aq = self.__size(a) z = [e for e in xrange(1, q+1)] tt = np.setdiff1d(z, a) tt_ind = np.setdiff1d(z, tt)-1 if x.ndim == 1: x = x[tt_ind] else: x = x[:, tt_ind] y = x v = [codea(aq, y, 10) for i in range(0, p)] if y.ndim == 1 \ else [codea(aq, y[i, :], 10) for i in range(0, p)] y = np.transpose(v) if y.shape[0] == 1 and len(y.shape) == 1: I, yy = [1], [y] y = np.hstack((y, I)) b, k, l = [y], [1], [1] else: ax = 1 if y.ndim > 1 else 0 yy = np.sort(y, axis=ax) I = y.argsort(axis=ax) y = np.hstack((yy, I)) b, k, l = np.unique(yy, return_index=True, return_inverse=True) y = np.hstack((l, I)) m = np.max(l) aa = np.zeros((m+1, p), dtype=int) for ii in range(0, m+1): for j in range(0, p): if l[j] == ii: aa[ii, j] = I[j]+1 return aa ''' Calculates lower approximation set of y ''' def rslower(self, y, a, T): z = self.indisc(a, T) w = [] p, q = self.__size(z) for u in range(0, p): zz = np.setdiff1d(z[u, :], 0) if np.in1d(zz, y).all(): w = np.hstack((w, zz)) return w.astype(dtype=int) ''' Calculates upper approximation set of y ''' def rsupper(self, y, a, T): z = self.indisc(a, T) w = [] p, q = self.__size(z) for u in range(0, p): zz = np.setdiff1d(z[u, :], 0) zzz = np.intersect1d(zz, y) if len(zzz) > 0: w = np.hstack((w, zz)) return w.astype(dtype=int) def __pospq(self, p, q): pm, pn = self.__size(p) qm, qn = self.__size(q) num = 0 pp, qq = [[]] * pm, [[]] * qm for i in range(0, pm): pp[i] = np.unique(p[i, :]) for j in range(0, qm): qq[j] = np.unique(q[j, :]) b = [] for i in range(0, qm): for j in range(0, pm): if np.in1d(pp[j], qq[i]).all(): num += 1 b = np.hstack((b, pp[j])) bb = np.unique(b) if bb.size == 0: dd = 1 else: _, dd = self.__size(bb) y = float(dd - 1)/pn if 0 in bb else float(dd)/pn b = np.setdiff1d(bb, 0) return y, b ''' Extract core set from C to D ''' def core(self, C, D): x = np.hstack((C, D)) c = np.array(range(1, C.shape[1]+1)) d = np.array([C.shape[1]+1]) cp, cq = self.__size(c) q = self.indisc(d, x) pp = self.indisc(c, x) b, w = self.__pospq(pp, q) a, k, kk, p = ([[]] * cq for i in range(4)) y = [] for u in range(0, cq): ind = u+1 a[u] = np.setdiff1d(c, ind) p[u] = self.indisc(a[u], x) k[u], kk[u] = self.__pospq(p[u], q) if k[u] != b: y = np.hstack((y, ind)) return np.array(y) def __sgf(self, a, r, d, x): pr = self.indisc(r, x) q = self.indisc(d, x) b = np.hstack((r, a)) pb = self.indisc(b, x) p1, _ = self.__pospq(pb, q) p2, _ = self.__pospq(pr, q) return p1 - p2 ''' Return the set of irreducible attributes ''' def reduce(self, C, D): def redu2(i, re, c, d, x): yre = re re1, re2 = self.__size(re) q = self.indisc(d, x) p = self.indisc(c, x) pos_cd, _ = self.__pospq(p, q) y, j = None, None for qi in range(i, re2): re = np.setdiff1d(re, re[qi]) red = self.indisc(re, x) pos_red, _ = self.__pospq(red, q) if np.array_equal(pos_cd, pos_red): y = re j = i break else: y = yre j = i + 1 break return y, j x = np.hstack((C, D)) c = np.array(range(1, C.shape[1]+1)) d = np.array([C.shape[1]+1]) y = self.core(C, D) q = self.indisc(d, x) p = self.indisc(c, x) pos_cd, _ = self.__pospq(p, q) re = y red = self.indisc(y, x) pos_red, _ = self.__pospq(red, q) while pos_cd != pos_red: cc = np.setdiff1d(c, re) c1, c2 = self.__size(cc) yy = [0] * c2 for i in range(0, c2): yy[i] = self.__sgf(cc[i], re, d, x) cd = np.setdiff1d(c, y) d1, d2 = self.__size(cd) for i in range(d2, c2, -1): yy[i] = [] ii = np.argsort(yy) for v1 in range(c2-1, -1, -1): v2 = ii[v1] re = np.hstack((re, cc[v2])) red = self.indisc(re, x) pos_red, _ = self.__pospq(red, q) re1, re2 = self.__size(re) core = y for qi in range(re2-1, -1, -1): if re[qi] in core: y = re break re = np.setdiff1d(re, re[qi]) red = self.indisc(re, x) pos_red, _ = self.__pospq(red, q) if np.array_equal(pos_cd, pos_red): y = re y1, y2 = self.__size(y) j = 0 for i in range(0, y2): y, j = redu2(j, y, c, d, x) return y #reduction属相约简 import numpy as np from sklearn.base import BaseEstimator from sklearn.feature_selection.base import SelectorMixin from scikit_roughsets.roughsets import RoughSetsReducer class RoughSetsSelector(BaseEstimator, SelectorMixin): def _get_support_mask(self): return self.mask_ def fit(self, X, y=None): # Missing values are not supported yet! if np.isnan(X).any(): raise ValueError("X must not contain any missing values") if np.isnan(y).any(): raise ValueError("y must not contain any missing values") # Check that X and Y contains only integer values if not np.all(np.equal(np.mod(X, 1), 0)): raise ValueError("X must contain only integer values") if not np.all(np.equal(np.mod(y, 1), 0)): raise ValueError("y must contain only integer values") reducer = RoughSetsReducer() selected_ = reducer.reduce(X, y) B_unique_sorted, B_idx = np.unique(np.array(range(X.shape[1])), return_index=True) B_unique_sorted = B_unique_sorted + 1 # Shift elements by one, as RS index array starts by one self.mask_ = np.in1d(B_unique_sorted, selected_, assume_unique=True) if self.mask_.size == 0: raise ValueError("No features were selected by rough sets reducer") return self