Python实战：用中智集解决模糊决策问题（附完整代码）
在数据科学和机器学习领域，决策问题往往伴随着不确定性。传统的模糊集理论已经无法完全满足复杂场景下的需求，这时中智集（Neutrosophic Set）作为一种更强大的数学工具应运而生。它不仅考虑隶属度和非隶属度，还引入了不确定性度量，为处理现实世界中的模糊决策问题提供了更精细的框架。
本文将带你深入理解中智集的核心概念，并通过Python实战演示如何将其应用于用户画像、风险评估等实际场景。我们将从单值中智集的基础实现开始，逐步扩展到区间中智集，最后探讨与scikit-learn的集成方法和可视化技巧。
1. 中智集基础与Python实现
1.1 理解中智集的三个维度
中智集的核心在于三个关键函数：
- 真隶属度(TA): 元素属于集合的程度
- 假隶属度(FA): 元素不属于集合的程度
- 不确定度(IA): 元素属于集合的不确定程度
这三个值满足：0 ≤ TA + IA + FA ≤ 3
让我们用Python定义一个基础的中智集类：
class SingleValuedNeutrosophicSet: def __init__(self, ta, ia, fa): if not (0 <= ta + ia + fa <= 3): raise ValueError("TA + IA + FA must be in [0,3]") self.ta = ta self.ia = ia self.fa = fa def __str__(self): return f"<{self.ta}, {self.ia}, {self.fa}>" def complement(self): return SingleValuedNeutrosophicSet(self.fa, self.ia, self.ta)1.2 单值中智集的基本运算
中智集支持多种运算，以下是Python实现的关键操作：
def union(s1, s2): ta = max(s1.ta, s2.ta) ia = min(s1.ia, s2.ia) fa = min(s1.fa, s2.fa) return SingleValuedNeutrosophicSet(ta, ia, fa) def intersection(s1, s2): ta = min(s1.ta, s2.ta) ia = max(s1.ia, s2.ia) fa = max(s1.fa, s2.fa) return SingleValuedNeutrosophicSet(ta, ia, fa) def score(s): """得分函数，用于比较中智数""" return (s.ta + 1 - s.ia + 1 - s.fa) / 32. 区间中智集的扩展实现
区间中智集将单值扩展为区间，提供了更大的灵活性。以下是Python实现：
class IntervalNeutrosophicSet: def __init__(self, ta_range, ia_range, fa_range): self.ta_low, self.ta_high = ta_range self.ia_low, self.ia_high = ia_range self.fa_low, self.fa_high = fa_range if not (0 <= self.ta_high + self.ia_high + self.fa_high <= 3): raise ValueError("TA + IA + FA must be in [0,3]") def __str__(self): return f"<[{self.ta_low},{self.ta_high}], [{self.ia_low},{self.ia_high}], [{self.fa_low},{self.fa_high}]>" def to_single_valued(self, alpha=0.5): """将区间中智集转换为单值中智集""" ta = alpha * self.ta_low + (1-alpha) * self.ta_high ia = alpha * self.ia_low + (1-alpha) * self.ia_high fa = alpha * self.fa_low + (1-alpha) * self.fa_high return SingleValuedNeutrosophicSet(ta, ia, fa)3. 与scikit-learn的集成
为了使中智集能够无缝融入现有的机器学习流程，我们可以创建一个兼容scikit-learn的转换器：
from sklearn.base import BaseEstimator, TransformerMixin class NeutrosophicTransformer(BaseEstimator, TransformerMixin): def __init__(self, membership_func, indeterminacy_func, non_membership_func): self.membership_func = membership_func self.indeterminacy_func = indeterminacy_func self.non_membership_func = non_membership_func def fit(self, X, y=None): return self def transform(self, X): result = [] for x in X: ta = self.membership_func(x) ia = self.indeterminacy_func(x) fa = self.non_membership_func(x) result.append([ta, ia, fa]) return np.array(result)使用示例：
def age_membership(age): if age < 20: return 1.0 elif age < 30: return 0.9 elif age < 40: return 0.7 else: return 0.3 def age_indeterminacy(age): if age < 25: return 0.1 elif age < 35: return 0.3 else: return 0.2 def age_non_membership(age): if age < 25: return 0.0 elif age < 35: return 0.2 else: return 0.5 transformer = NeutrosophicTransformer(age_membership, age_indeterminacy, age_non_membership) ages = np.array([[22], [28], [45]]) neutrosophic_features = transformer.transform(ages)4. 可视化技巧
中智集的三维特性使其可视化具有挑战性。以下是几种有效的可视化方法：
4.1 三元图表示
import matplotlib.pyplot as plt from mpl_toolkits.mplot3d import Axes3D def plot_neutrosophic_3d(points): fig = plt.figure(figsize=(10, 8)) ax = fig.add_subplot(111, projection='3d') for p in points: ax.scatter(p.ta, p.ia, p.fa) ax.set_xlabel('Truth Membership (TA)') ax.set_ylabel('Indeterminacy (IA)') ax.set_zlabel('False Membership (FA)') ax.set_title('Neutrosophic Set Visualization') plt.show()4.2 雷达图比较
def plot_neutrosophic_radar(labels, sets): categories = ['TA', 'IA', 'FA'] N = len(categories) angles = [n / float(N) * 2 * np.pi for n in range(N)] angles += angles[:1] fig = plt.figure(figsize=(8, 8)) ax = fig.add_subplot(111, polar=True) ax.set_theta_offset(np.pi / 2) ax.set_theta_direction(-1) plt.xticks(angles[:-1], categories) ax.set_rlabel_position(0) plt.yticks([0.2, 0.4, 0.6, 0.8, 1.0], ["0.2", "0.4", "0.6", "0.8", "1.0"], color="grey", size=7) plt.ylim(0, 1) for i, s in enumerate(sets): values = [s.ta, s.ia, s.fa] values += values[:1] ax.plot(angles, values, linewidth=1, linestyle='solid', label=labels[i]) ax.fill(angles, values, alpha=0.1) plt.legend(loc='upper right', bbox_to_anchor=(0.1, 0.1)) plt.show()5. 实际应用案例：用户画像
让我们通过一个完整的用户画像案例来展示中智集的实际应用：
class UserProfile: def __init__(self): self.categories = { 'tech_savvy': None, 'price_sensitive': None, 'brand_loyal': None } def add_category(self, name, neutrosophic_set): self.categories[name] = neutrosophic_set def similarity(self, other): """计算两个用户画像的相似度""" total = 0 for cat in self.categories: s1 = self.categories[cat] s2 = other.categories[cat] # 使用余弦相似度变体 numerator = s1.ta*s2.ta + s1.ia*s2.ia + s1.fa*s2.fa denominator = (s1.ta**2 + s1.ia**2 + s1.fa**2)**0.5 * \ (s2.ta**2 + s2.ia**2 + s2.fa**2)**0.5 total += numerator / denominator return total / len(self.categories) # 创建用户画像 user1 = UserProfile() user1.add_category('tech_savvy', SingleValuedNeutrosophicSet(0.9, 0.1, 0.1)) user1.add_category('price_sensitive', SingleValuedNeutrosophicSet(0.3, 0.2, 0.6)) user1.add_category('brand_loyal', SingleValuedNeutrosophicSet(0.7, 0.3, 0.2)) user2 = UserProfile() user2.add_category('tech_savvy', SingleValuedNeutrosophicSet(0.8, 0.2, 0.2)) user2.add_category('price_sensitive', SingleValuedNeutrosophicSet(0.4, 0.3, 0.5)) user2.add_category('brand_loyal', SingleValuedNeutrosophicSet(0.6, 0.4, 0.3)) print(f"用户相似度: {user1.similarity(user2):.2f}")6. 性能优化与高级技巧
当处理大规模数据时，我们需要考虑性能优化：
import numpy as np from numba import njit @njit def fast_neutrosophic_operation(ta1, ia1, fa1, ta2, ia2, fa2, operation='union'): if operation == 'union': ta = max(ta1, ta2) ia = min(ia1, ia2) fa = min(fa1, fa2) elif operation == 'intersection': ta = min(ta1, ta2) ia = max(ia1, ia2) fa = max(fa1, fa2) return ta, ia, fa # 批量处理示例 def batch_operation(set1, set2, operation='union'): ta1 = np.array([s.ta for s in set1]) ia1 = np.array([s.ia for s in set1]) fa1 = np.array([s.fa for s in set1]) ta2 = np.array([s.ta for s in set2]) ia2 = np.array([s.ia for s in set2]) fa2 = np.array([s.fa for s in set2]) results = [] for t1, i1, f1, t2, i2, f2 in zip(ta1, ia1, fa1, ta2, ia2, fa2): ta, ia, fa = fast_neutrosophic_operation(t1, i1, f1, t2, i2, f2, operation) results.append(SingleValuedNeutrosophicSet(ta, ia, fa)) return results在实际项目中，我发现将中智集与传统的机器学习算法结合时，特征工程阶段最为关键。通常需要设计特定的转换函数，将中智集的三个维度合理地映射到模型可理解的特征空间。
:百万级数据 MySQL 查询优化宝典)