回归树，在决策书节点选择计算MSE/方差降低，在叶子节点选择均值。
| 487 | |
| 488 | |
| 489 | class RegressionTree(DecisionTree): |
| 490 | """ |
| 491 | 回归树，在决策书节点选择计算MSE/方差降低，在叶子节点选择均值。 |
| 492 | """ |
| 493 | def _calculate_mse(self, y, y1, y2): |
| 494 | """ |
| 495 | 计算MSE降低 |
| 496 | """ |
| 497 | mse_tot = calculate_mse(y) |
| 498 | mse_1 = calculate_mse(y1) |
| 499 | mse_2 = calculate_mse(y2) |
| 500 | frac_1 = len(y1) / len(y) |
| 501 | frac_2 = len(y2) / len(y) |
| 502 | mse_reduction = mse_tot - (frac_1 * mse_1 + frac_2 * mse_2) |
| 503 | return mse_reduction |
| 504 | |
| 505 | def _calculate_variance_reduction(self, y, y1, y2): |
| 506 | """ |
| 507 | 计算方差降低 |
| 508 | """ |
| 509 | var_tot = calculate_variance(y) |
| 510 | var_1 = calculate_variance(y1) |
| 511 | var_2 = calculate_variance(y2) |
| 512 | frac_1 = len(y1) / len(y) |
| 513 | frac_2 = len(y2) / len(y) |
| 514 | variance_reduction = var_tot - (frac_1 * var_1 + frac_2 * var_2) |
| 515 | return sum(variance_reduction) |
| 516 | |
| 517 | def _mean_of_y(self, y): |
| 518 | """ |
| 519 | 计算均值 |
| 520 | """ |
| 521 | value = np.mean(y, axis=0) |
| 522 | return value if len(value) > 1 else value[0] |
| 523 | |
| 524 | def fit(self, X, y): |
| 525 | self._impurity_calculation = self._calculate_mse |
| 526 | self._leaf_value_calculation = self._mean_of_y |
| 527 | super(RegressionTree, self).fit(X, y) |
| 528 | |
| 529 | |
| 530 | ########-----PCA------######### |