使用python写一段smote算法

Sure! 下面是一个基本的SMOTE算法的Python代码：

import numpy as np
from sklearn.neighbors import NearestNeighbors

def smote(X, y, k=5, ratio=1):
    """
    Generate synthetic samples using SMOTE algorithm.
    :param X: 2D numpy array of shape (n_samples, n_features)
    :param y: 1D numpy array of shape (n_samples,), the corresponding class labels
    :param k: int, the number of nearest neighbors to use when selecting similar samples
    :param ratio: float, the amount of SMOTE to apply (e.g. 0.5 means increase the minority class by 50%)
    :return: 2D numpy array of shape (n_samples', n_features), the synthetic samples
    """
    # Get the unique class labels and the number of samples in each class
    classes, counts = np.unique(y, return_counts=True)
    # Find the minority class
    minority_class = classes[np.argmin(counts)]
    # Find the indices of the minority class samples
    minority_indices = np.where(y == minority_class)[0]
    # Calculate the number of synthetic samples to generate
    n_to_generate = int(ratio * counts[np.argmin(counts)])
    # Initialize an empty array to hold the synthetic samples
    synthetic_samples = np.zeros((n_to_generate, X.shape[1]))
    # Fit a k-NN model to the original data
    knn = NearestNeighbors(n_neighbors=k).fit(X)
    # Generate the synthetic samples
    for i in range(n_to_generate):
        # Choose a random minority class sample
        idx = np.random.choice(minority_indices)
        # Find its k nearest neighbors in the original data
        nn = knn.kneighbors(X[idx].reshape(1, -1), return_distance=False)[0]
        # Choose one of the neighbors randomly
        nn_idx = np.random.choice(nn)
        # Calculate the difference between the minority sample and the neighbor
        diff = X[nn_idx] - X[idx]
        # Multiply this difference by a random value between 0 and 1
        gap = np.random.rand() * diff
        # Add this gap to the minority sample to create the synthetic sample
        synthetic_samples[i, :] = X[idx] + gap
    # Combine the original data with the synthetic data
    new_X = np.vstack((X, synthetic_samples))
    new_y = np.hstack((y, np.array([minority_class] * n_to_generate)))
    # Shuffle the data and return it
    idx = np.random.permutation(new_X.shape[0])
    return new_X[idx], new_y[idx]

使用方法：

# 载入数据
X, y = load_data()
# 使用 SMOTE 生成新的样本
X_smote, y_smote = smote(X, y, k=5, ratio=0.5)