### 使用 Cosine Similarity 计算 Two Groups 的 Feature Difference 为了计算两个群体之间的特征差异，可以采用余弦相似度（Cosine Similarity）。该方法通过测量两个向量之间夹角的余弦值来评估它们的方向上的相似程度。当处理高维空间中的数据时，这种方法尤其有效。 下面是一个具体的实现方式： #### 准备工作 首先安装必要的库 `scikit-learn` 和 `numpy`: ```bash pip install scikit-learn numpy ``` #### 编写代码 接着编写一段用于计算两组样本间平均余弦相似度得分的 Python 代码如下所示： ```python from sklearn.metrics.pairwise import cosine_similarity import numpy as np def calculate_group_difference(group_a_features, group_b_features): """ Calculate average cosine similarity score between two groups. Parameters: group_a_features (list/array): Features matrix for Group A samples. group_b_features (list/array): Features matrix for Group B samples. Returns: float: Average cosine similarity across all pairs of instances from both groups. """ # Convert lists to NumPy arrays if they aren't already features_A = np.array(group_a_features) features_B = np.array(group_b_features) # Compute pairwise cosine similarities sim_matrix = cosine_similarity(features_A, features_B) # Return mean value over entire matrix return sim_matrix.mean() # Example usage if __name__ == "__main__": # Define some example feature vectors for demonstration purposes only group_1_samples = [[0.8, 0.2], [0.9, 0.1]] group_2_samples = [[0.3, 0.7], [0.4, 0.6]] avg_sim_score = calculate_group_difference(group_1_samples, group_2_samples) print(f'Average Cosine Similarity Score Between Groups: {avg_sim_score:.4f}') ``` 这段程序定义了一个函数 `calculate_group_difference()` 来接收来自不同群集的数据点作为输入参数，并返回这些群集中所有可能配对实例间的平均余弦相似度分数[^1]。