尽可能快地运行随机森林模型的最佳实践

Question

Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.

---

这个问题似乎不是关于在

闭合 2年前。

Answer 1

1. 随机子采样，然后在完整数据上进行调整很少有效，因为小的子样本可能不能代表完整数据。

2. 关于数据的数量与模型质量：尝试使用Sklearn的学习曲线： https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.learning_curve.html

    train_sizes, train_scores, test_scores, fit_times, _ = learning_curve(
        estimator,
        X,
        y,
        cv=cv,
        n_jobs=n_jobs,
        train_sizes=train_sizes,
        return_times=True,
    )

您将能够绘制数据与模型性能的量。
以下是绘图的一些示例：
https://scikit-learn.org/stable/auto_examples/miscellaneous/plot_kernel_ridge_ridge_regression.html#sphx-glr-auto-auto-auto-auto-auto-examples-miscellaneous-miscellaneous-plot-kernel-ridge-ridge-ridge-ridge-ridge-ridge-ridge-py-py-py-py-py-py-py-ridge-ridge-py-py

“ https://scikit-learn.org/stable/auto_examples/model_selection/plot_learning_curve.html#sphx-glr-auto-auto-auto-examples-model-selection-selection-plot-selection-plot-learnning-learnning-learnning-curve-curve-curve-py-py-py-curve-py-curve-py-curve-curve-py- /scikit-learn.org/stable/auto_examples/model_selection/plot_learning_curve.html#sphx-glr-auto-auto-examples-model-selection-selection-plot-learning-learning-curve-curve-curve-curve-curve-curve-curve-py-py

3. 估计总时间是困难的，因为它很难T线性。

一些其他实用建议：

• SET n_jobs = -1以在所有内核上并行运行该模型；
• 使用任何功能选择方法来减少功能数量。 300个功能确实有很多，应该可以摆脱大约一半的功能，而不会大幅下降模型性能。