"""
用到函数库： pandas, numpy, openpyxl, matplotlib
预测不同的网络要选用不同的参数：pre_number ：向后预测的次数,每次增加一个值use_end_number ：每次预测所用末尾数据的条目数"""import pandas as pd
import numpy as np
import openpyxlimport matplotlib.pyplot as plt
import matplotlib.pyplot
matplotlib.rcParams['font.sans-serif'] = ['FangSong'] # 指定默认字体
matplotlib.rcParams['axes.unicode_minus'] = False # 解决保存图像是负号'-'显示为方块的问题class GrayForecast():def __init__(self, data, datacolumn=None):if isinstance(data, pd.core.frame.DataFrame):self.data=datatry:self.data.columns = ['数据']except:if not datacolumn:raise Exception('您传入的dataframe不止一列')else:self.data = pd.DataFrame(data[datacolumn])self.data.columns=['数据']elif isinstance(data, pd.core.series.Series):self.data = pd.DataFrame(data, columns=['数据'])else:self.data = pd.DataFrame(data, columns=['数据'])self.forecast_list = self.data.copy()if datacolumn:self.datacolumn = datacolumnelse:self.datacolumn = None#save arg:#        data                DataFrame    数据#        forecast_list       DataFrame    预测序列#        datacolumn          string       数据的含义# 数据级比校验def level_check(self):n = len(self.data)lambda_k = np.zeros(n-1)for i in range(n-1):lambda_k[i] = self.data[i]["数据"]/self.data[i+1]["数据"]if lambda_k[i] < np.exp(-2/(n+1)) or lambda_k[i] > np.exp(2/(n+2)):flag = Falseelse:flag = Trueself.lambda_k = lambda_kif not flag:print("级比校验失败，请对X(0)做平移变换")return Falseelse:print("级比校验成功，请继续")return True#save arg:#        lambda_k            1-d list# 灰色预测为短期预测，过多的数据反而会导致数据精准度变差
# 参数forecast为使用forecast_list末尾数据的数量def GM_11_build_model(self, forecast=5):if forecast > len(self.data):raise Exception('您的数据行不够')X_0 = np.array(self.forecast_list['数据'].tail(forecast))# print(self.forecast_list)# print(X_0)#       1-AGOX_1 = np.zeros(X_0.shape)for i in range(X_0.shape[0]):X_1[i] = np.sum(X_0[0:i+1])#       紧邻均值生成序列Z_1 = np.zeros(X_1.shape[0]-1)for i in range(1, X_1.shape[0]):Z_1[i-1] = -0.5*(X_1[i]+X_1[i-1])B = np.append(np.array(np.mat(Z_1).T), np.ones(Z_1.shape).reshape((Z_1.shape[0], 1)), axis=1)Yn = X_0[1:].reshape((X_0[1:].shape[0], 1))B = np.mat(B)Yn = np.mat(Yn)a_ = (B.T*B)**-1 * B.T * Yna, b = np.array(a_.T)[0]X_ = np.zeros(X_0.shape[0])def f(k):return (X_0[0]-b/a)*(1-np.exp(a))*np.exp(-a*(k))self.forecast_list.loc[len(self.forecast_list)] = f(X_.shape[0])# print(self.forecast_list)# 参数time为向后预测的次数,每次增加一个值，forecast_data_len为每次预测所用末尾数据的条目数def forecast(self, time=5, forecast_data_len=5):for i in range(time):self.GM_11_build_model(forecast=forecast_data_len)print('预测结果:', self.forecast_list[-time:])           # def log(self):#     res = self.forecast_list.copy()#     if self.datacolumn:#         res.columns = [self.datacolumn]#     return res   
# 初始化序列def reset(self):self.forecast_list = self.data.copy()def plot(self,x, fig, pre_number):# self.forecast_list.plot()y = self.forecast_list.copy()[-pre_number:]# print(y)# print(len(self.data))a = list()b = len(x)# for i in range(0, pre_number):#     a.append(b+i)# print(a)for i in range(0, pre_number):i = pre_number - ia.append(b-i)print(a)fig.plot(a, y,'o-')# fig.plot(x[-pre_number:], y,'o-')# plt.title('预测数据',loc ='left',color='red')# plt.ylabel(self.datacolumn)# plt.legend([self.datacolumn])# plt.imshow()plt.show()if __name__ == "__main__":#将excel数据提取出来，并转成list，将data第一行汉字删除 def get_df(colum):wb = openpyxl.load_workbook('data1.xlsx')# wb = openpyxl.load_workbook('NorthAmerica.xlsx')ws = wb.active  # 当前活跃的表单G = ws[colum]df = []for cell in G:df.append(cell.value)df = list(filter(None, df))# print('原始数据 - 测试数据：', len(df))# print('原始数据', df)return dfdef plot(col):ax1 = plt.subplot(111)df2 = get_df(col)x = [i for i in range(0, len(df2))]print('原始数据个数:', len(df2))ax1.plot(x, df2, 'x--')plt.title('尾部不同颜色为预测值：',loc ='left',color='red')plt.show()return x, ax1# 传入数据，每次一列df = get_df('D')x, ax1 = plot('E')print('原始数据 - 测试数据：', len(df))    # 向后预测的次数,每次增加一个值# 每次预测所用末尾数据的条目数# 根据列数据的多少，会有调整pre_number = 2use_end_number = 8gf = GrayForecast(df)gf.forecast(pre_number,use_end_number)gf.plot(x, ax1, pre_number)