DBSCAN调参_

''' DBSCAND-1 此程序是为了DBSCAND调参,寻找合适的 esp 和 min_samples 并将生成的各类参数组合和聚类结果导出到csv文件。 ''' import numpy as np import matplotlib.pyplot as plt from sklearn import cluster import pandas as pd import os time = '19年_11月' os.mkdir('%s聚类结果图' % time) # 获取上一级(父)路径 parent_path = os.path.dirname(os.path.dirname(__file__)) path = parent_path + '/Setp-2-根据时间对选出的数据聚合/降采样-1min/' data = pd.read_csv(path+'084号_%s_40项原始数据_1min.csv' % time) data = data.dropna(axis=0,how='any') #去除null值 X = data[['环境风速', '有功功率']].values res = [] '''迭代eps范围,三个参数分别为(起始,结束,步长)''' for eps in np.arange(0.501, 3, 0.05): '''迭代min_samples,两个参数分别为(起始,结束)，步长默认为1''' for min_samples in range(2, 16): dbscan = cluster.DBSCAN(eps=eps, min_samples=min_samples) dbscan.fit(X) # 新增一列dbscan_label 为分类标签 data['dbscan_label'] = dbscan.fit_predict(X) normal = data[data['dbscan_label'] != -1] unnormal = data[data['dbscan_label'] == -1] normalX, normalY = normal['环境风速'], normal['有功功率'] unnormalX, unnormalY = unnormal['环境风速'], unnormal['有功功率'] plt.rcParams['font.sans-serif'] = ['SimHei'] # 显示中文标签 plt.scatter(normalX, normalY, marker='.', color='g', label='正常值') plt.scatter(unnormalX, unnormalY, marker='+', color='r', label='噪声值') plt.xlabel('风速(m/s)') plt.ylabel('功率(Kw)') plt.title('No84-%s-功率和风速关系散点图,esp=%.3f min_samples = %d' % (time, eps, min_samples)) plt.legend() # 显示图中label plt.savefig('%s聚类结果图/esp=%.3f-min_samples=%d.png' % (time, eps, min_samples)) plt.show() # 统计各参数组合下的聚类个数 n_clusters（-1表示异常点） n_clusters = len([i for i in set(dbscan.labels_) if i != - 1]) # 异常点的个数 outliners outliners = np.sum(np.where(dbscan.labels_ == - 1, 1, 0)) # 统计每个簇的样本个数 stats stats = str(pd.Series([i for i in dbscan.labels_ if i != - 1]).value_counts().values) res.append({'eps': eps, 'min_samples': min_samples, 'n_clusters/聚类个数': n_clusters, 'outliners/异常点个数': outliners}) df = pd.DataFrame(res) df = pd.DataFrame(res) print(res) df.to_csv('%s聚类结果图/DBSCAND%s.csv' % (time, time), index=False,encoding='utf_8_sig')