二次指数平滑算法Python实现

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import pymysql

# Read data from csv files

# df = pd.read_csv('E://CapacityData/test-data-set/MeishanRenshou.csv')

# Read data from mysql, including downstream_userplane_flux, max_rrc_connection_number.

conn = pymysql.connect(user='root', passwd='root123',

                       host='localhost', db='capacitypredict', charset='utf8')

cur = conn.cursor()

cur.execute("select concat(netelement,cellid)as id , max_rrc_connection_number from table"

            "where DATE_FORMAT(starttime, '%Y-%m-%d')>='2016-11-01' "

            "and DATE_FORMAT(starttime, '%Y-%m-%d')<='2016-12-07'"

            "order by id ")

sqlData = cur.fetchall()

cur.close()

conn.close()

print(sqlData)

originalId = sqlData[0][0]

data = []

SRMSE = []; SMAPE = []; SMSE = []; fPredictData = []; fOriginalData =[]

lead = 7  # Define predict steps

# Define alpha, which used to adjust weights

alpha = 0.3

for d in sqlData:

    print(d)

    if d[0] == originalId:

        data.append(float(d[1]))

    else:

        # Testing the data's length

        if len(data) == 37:

            # In this section, we calculate the one-ES

            S1 = []

            firstS1 = 0

            for t in range(0, len(data)):

                if t == 0:

                    for n in range(0, 3):

                        print(data)

                        firstS1 += float(data
)

                    S1.append(firstS1/3)

                else:

                    S1_diff = alpha*data[t-1] + (1-alpha)*S1[t-1]

                    S1.append(S1_diff)

            print(S1)

            # In the next part, we compute the two—ES

            S2 = []

            firstS2 = 0

            for s in range(0, len(S1)):

                if s == 0:

                    for m in range(0, 3):

                        firstS2 += float(S1[m])

                    S2.append(firstS2/3)

                else:

                    S2_diff = alpha*S1[s-1] + (1-alpha)*S2[s-1]

                    S2.append(S2_diff)

            print(S2[1:10]) # Output several S2 data

            # Next, calculate the parameters slope & intercept

            slope = 0

            intercept = 0

            predictValues = []

            for i in range(0, len(data)):

                intercept = 2*S1[i] - S2[i]

                slope = (alpha/(1-alpha))*(S1[i]-S2[i])

                predict = intercept + slope*lead

                predictValues.append(abs(predict))

            # Define how many days or weeks, we want to predict: K_count

            K_count = 7

            # original data

            originalData = data[len(data)-K_count:len(data)]

            print(len(originalData))

            # predicted values. I discard the last value, because it is the predicted value for t+1

            predictData = predictValues[len(predictValues) - lead - K_count:len(predictValues) - lead]

            print(len(predictData))

            # Computing the indicators like RMSE, MAPE //(import math math.sqrt)

            MSE = sum((np.array(originalData)-predictData)**2)/K_count

            MAPE = sum(abs(np.array(originalData)-predictData)/(np.array(originalData)+0.001))/K_count

            print("the model's RMSE is: ", MSE)

            print("the model's MAPE is: ", MAPE)

            SMSE.append(MSE)

            SMAPE.append(MAPE)

            # Resetting the var: data

            data = []

            # Alter the originalId

            print(d)

            originalId = d[0]

            data.append(float(d[1]))

            # Storing the predicted data and

            fOriginalData.append(originalData)

            fPredictData.append(predictData)

        else:

            data = []

            originalId = d[0]

            data.append(float(d[1]))

# Print the final values of error indicators

meanMse = sum(SMSE) / len(SMSE)
print("This district's mean mse is: ", meanMse)