雨型K均值聚类

# -*- coding: utf-8 -*-
"""
@Author: xxx
@Features of this code:利用聚类找出24、48、72小时雨型
"""
import numpy as np
import pandas as pd
import glob
# import matplotlib.pyplot as plt
from sklearn.cluster import KMeans  # 导入K均值聚类算法


# from pylab import *


# 读取数据
def read_data(filename):
    f = open(filename, 'r')
    file = f.readlines()
    temp = []
    for r in range(len(file)):
        line = file[r].split()
        temp = pd.concat([pd.DataFrame(temp), pd.DataFrame([line])], axis=0)
    f.close()
    data_fun = np.array(temp).astype(float)
    data_fun = np.hstack([np.zeros([30, 1]), data_fun])
    data_fun = np.delete(data_fun, -1, 1)  # 删除最后一列
    return data_fun


# 处理成各小时雨量分布
def get_pre_h(data, points):
    data_h = np.zeros([np.size(data, 0), points])
    data_h1 = np.zeros([np.size(data, 0), points])
    for r in range(np.size(data, 0)):
        data_line = np.delete(data[r, 4::], np.isnan(data[r, 4::]))  # 短历时从第3列开始
        delta = np.floor(len(data_line) / points).astype(int)
        for i in range(points):
            data_h[r, i] = np.sum(data_line[i * delta: (i + 1) * delta])
        for i in range(points):
            data_h1[r, i] = data_h[r, i] / np.sum(data_h[r, :])
    return data_h1


# K聚类分析
def k_m(h, k):
    # k 为类数
    k_model = KMeans(n_clusters=k)  # 调用K聚类算法
    k_model.fit(h)  # 训练模型
    k_count = pd.Series(k_model.labels_).value_counts()  # 统计各个类别的数目
    k_center = pd.DataFrame(k_model.cluster_centers_)  # 找出聚类中心
    k_cst1 = pd.concat([k_count, k_center], axis=1)  # 横向连接（0是纵向），得到聚类中心对应的类别下的数目
    k_cst2 = pd.concat([pd.Series(k_model.labels_), pd.DataFrame(h)], axis=1)  # 详细输出每个样本对应的类别
    return k_cst1, k_cst2


if __name__ == '__main__':
    files = glob.glob(r'*.txt')  # 批量读取数据
    for fn in files:
        data_input = read_data(fn)
        # hours_count = int(np.size(data_input, 1) / 12)
        # data_h = np.zeros([30, hours_count])
        # for iter_h in range(hours_count):
        #     data_h[:, iter_h] = np.sum(data_input[:, iter_h * 12:iter_h * 12 + 12], 1)
        cluster_count_center, cluster_type = k_m(data_input, 5)
        class_name = ['第1类', '第2类', '第3类', '第4类', '第5类']
        cluster_count_center.index = class_name
        # top30_1h
        cluster_count_center.to_excel(fn[0:48] + '聚类.xlsx')  ## top30_1h
        print(fn[0:48] + '聚类.xlsx' + '保存完毕！')
        # top30_2h
        # cluster_count_center.to_excel(fn[0:51] + '聚类.xlsx')
        # print(fn[0:51]+'聚类.xlsx'+'保存完毕！')