算法一类支持向量机OC-SVM

#abc-svm Demo1 import numpy as np from sklearn.datasets import make_blobs from sklearn.svm import OneClassSVM from sklearn.metrics import roc_auc_score from sklearn.model_selection import train_test_split # 蜂群算法类 class ArtificialBeeColony: def __init__(self, fitness_func, param_ranges,kernel_options, population_size=30, max_generations=100): self.fitness_func = fitness_func self.param_ranges = param_ranges self.kernel_options = kernel_options # 添加kernel选项 self.population_size = population_size self.max_generations = max_generations self.population = self.initialize_population() self.best_solution = None self.best_fitness = -np.inf def initialize_population(self): population = [] for _ in range(self.population_size): solution = {} for param_name, (low, high, step) in self.param_ranges.items(): if isinstance(low, list): # Handle categorical variables solution[param_name] = np.random.choice(low) else: solution[param_name] = low + step * np.random.rand() * (high - low) # 随机选择一个kernel solution['kernel'] = np.random.choice(self.kernel_options) population.append(solution) return population def update_population(self, new_solutions): self.population = new_solutions current_best = max(new_solutions, key=lambda x: self.fitness_func(x)) current_fitness = self.fitness_func(current_best) if current_fitness > self.best_fitness: self.best_fitness = current_fitness self.best_solution = current_best def optimize(self): for generation in range(self.max_generations): new_population = [] for solution in self.population: new_solution = self.search_for_new_solution(solution) new_population.append(new_solution) self.update_population(new_population) print(f"Generation {generation + 1}: Best Fitness = {self.best_fitness}") return self.best_solution, self.best_fitness def search_for_new_solution(self, solution): new_solution = {} for param_name, value in solution.items(): if param_name == 'kernel': # 对于kernel参数，从kernel_options列表中随机选择一个值 new_solution[param_name] = np.random.choice(self.kernel_options) else: low, high, step = self.param_ranges[param_name] if isinstance(low, list): # Handle categorical variables new_value = np.random.choice(low) else: new_value = value + step * (np.random.rand() - 0.5) * (high - low) new_value = max(low, min(new_value, high)) # Ensure value is within bounds new_solution[param_name] = new_value return new_solution # 创建模拟数据集 X, y = make_blobs(n_samples=500, centers=1, n_features=20, random_state=42) # print(X) # print("---------") # print(y) # print("---------") from sklearn.metrics import r2_score # 适应度函数 def fitness_function(params): # 选取X_train 为正类 X_train, X_test, _, y_test = train_test_split(X, y, test_size=0.3, random_state=42) y_test = np.where(y_test == 0, -1, 1) # 将标签转换为-1和1 nu = params['nu'] kernel = params['kernel'] gamma = params['gamma'] # 创建和训练One-Class SVM模型 ocsvm = OneClassSVM(nu=nu, kernel=kernel, gamma=gamma) ocsvm.fit(X_train) # ocsvm.fit(X) # 预测测试集并计算ROC AUC分数作为适应度值 y_pred = ocsvm.predict(X_test) # fitness = roc_auc_score(y_test, y_pred) # fitness = roc_auc_score(y, y_pred) fitness =r2_score(y_test, y_pred) return fitness # 参数搜索范围 param_ranges = { 'nu': (0.1, 0.9, 0.1), # nu 的搜索范围，步长为 0.1 # 'kernel': ['rbf', 'linear'], # 核函数的选择列表 'gamma': (0.01, 1, 0.1) # gamma 的搜索范围，步长为 0.1 } kernel_options = ['rbf', 'linear'] # 可能的kernel值列表 # 创建并运行人工蜂群算法 # abc = ArtificialBeeColony(fitness_function, param_ranges) abc = ArtificialBeeColony(fitness_function, param_ranges, kernel_options) best_solution, best_fitness = abc.optimize() # Best Solution: {'nu': 0.10587043848153328, 'gamma': 0.2575143995541765, 'kernel': 'rbf'} # Best Fitness: 1.0 # 输出最佳解和最佳适应度 print(f"Best Solution: {best_solution}") print(f"Best Fitness: {best_fitness}")