ARIMA 模型预测用户活跃情况

数据清洗，数据格式

import pandas as pd

user={}
with open("/home/kesci/input/com6636/user.txt",'r') as f:
    for line in f.readlines():
        num=line.split(' ')
        l=[]
        for i in range(1,len(num)):
            l.append(int(num[i].replace("\n",'')))
        user[int(num[0])]=l
        
#print(user)
#for key in user.keys():
    #print(key)
    #break
l={}
#for key in user.keys():
    #if key<=:
       # print(user[key])
x=[x for x in range(1,31)]
    #user=data
user=[float(x) for x in user[105]]
dict_1={"date":x,"user":user}
data=pd.DataFrame(dict_1)
data
date	user
0	1	0.0
1	2	0.0
2	3	0.0
3	4	0.0
4	5	0.0
5	6	0.0
6	7	0.0
7	8	0.0
8	9	0.0
9	10	0.0
10	11	0.0
11	12	2.0
12	13	0.0
13	14	2.0
14	15	2.0
15	16	1.0
16	17	2.0
17	18	2.0
18	19	2.0
19	20	2.0
20	21	2.0
21	22	2.0
22	23	2.0
23	24	0.0
24	25	2.0
25	26	0.0
26	27	1.0
27	28	2.0
28	29	0.0
29	30	2.0
# 显示cell运行时长

import pandas as pd
user={}
with open("/home/kesci/input/com6636/user.txt",'r') as f:
    for line in f.readlines():
        num=line.split(' ')
        l=[]
        for i in range(1,len(num)):
            l.append(int(num[i].replace("\n",'')))
        user[int(num[0])]=l
        
#print(user)
#for key in user.keys():
    #print(key)
    #break

#import pandas as pd


       # print(i)
import warnings
warnings.filterwarnings("ignore")
#for key in dict（list(data_reg)）:

from statsmodels.tsa.arima_model import ARIMA
import pandas as pd
def model_build(data,i,j):                 #建立模型
    x=[x for x in range(1,31)]
    user=data
    user=[float(x) for x in user]
    dict_1={"date":x,"user":user}
    data=pd.DataFrame(dict_1)
#data=data["user"].apply(lambda x:float(x))
    dg_1=data[(data['date']!=0)].groupby(['date'])['user'].sum()
    model = ARIMA(dg_1.values, order=(i,1,j))
    return model
def pro(l):                             #预测情况
    l=[-x for x in l]
    for i in range(7):
       if l[i]<0:
           l[i]=0
    return l
def acc(x,key_3):                          #测试精度
    li=[]
    k=0
    l=0
    for i in range(29):
        if x[i]>0.09:
            li.append(1.0)
            k=k+1
        else:
            li.append(0.0)
        if user[key_3][i]!=0:
            l=l+1
    #print(li)
    m=k/l
    return m
from statsmodels.tsa.arima_model import ARIMA
import matplotlib.pyplot as plt
#model=model = ARIMA(dg_1.values, order=(1, 1, 2))
f_name=[]
loss_u=[]
#global loss_key=0
def find_c(u,key_2):                       #寻找ARIMA模型合适的q p 值
    pram_l=[0,0]
    #pram_l[0]=1
    #pram_l[1]=4
    ac_1=0.0
    flag_1=0
    
    for x in range(1,10):
      for y in range(1,10):
         try:
            model=model_build(u,x,y)
            results_AR = model.fit(disp=-1)
            f_pred_1=results_AR.predict(30,36)  
            #print(f_pred_1)
            pram_l.append(x)
            pram_l.append(y)
            break
            results_AR = model.fit(disp=-1)
            pred=[ -x for x in results_AR.fittedvalues]
#pred=[ -x for x in results_AR.predict(0,37)]
            for i in range(29):
               if pred[i]<0:
                  pred[i]=0
            if ac_1<acc(pred,key_2):
                ac_1=acc(pred,key_2)
                pram_l.append(x)
                pram_l.append(y)
         except:
                if x==9 and y==9:
                    #loss_key=loss_key+1
                    #loss_u.append(key)
                    flag_1=key
                    #print(loss_key)
                    model=model_build(u,0,0)
                    results_AR = model.fit(disp=-1)
                    f_pred_1=results_AR.predict(30,36)
                    #f_pred_1=
                else:
                    print("wait for .....")
    return (f_pred_1,flag_1)
try:
  k=0
  bb=0
  for key in range(0,5000):
    #k=k+1
    if key in user.keys():                       #遍历所有用户，对每个用户进行预测
        #bb=bb+1     
        print("speed====="+str(key))
        #l=find_c(user[key],key)
        #print(l[-2])
        #print(l[-1])
        try:
           model=model_build(user[key],1,2)
           results_AR = model.fit(disp=-1)  

#fig, ax = plt.subplots(figsize=(15,6))
#ax.plot(dg_1.values)
     #pred=[ -x for x in results_AR.fittedvalues]
#pred=[ -x for x in results_AR.predict(0,37)]
     #for i in range(29):
       # if pred[i]<0:
         #   pred[i]=0
   #print(results_AR.fittedvalues)
   #print(pred)
#print(pred[0])
#fig, ax = plt.subplots(figsize=(15,6))
#ax.plot(pred)
   #print(results_AR.predict(30,36))
           f_pred=results_AR.predict(30,36)
           for name in pro(f_pred):
               if name>=0.09:
                   f_name.append(str(key))
                   break
        except:
             l,x=find_c(user[key],key)
             if x!=0:
                 loss_u.append(x)
                 print(len(loss_u))
             #print(l[-2])
             #print(l[-1])
             #model=model_build(user[key],l[-2],l[-1])
             #results_AR = model.fit(disp=-1)  

#fig, ax = plt.subplots(figsize=(15,6))
#ax.plot(dg_1.values)
     #pred=[ -x for x in results_AR.fittedvalues]
#pred=[ -x for x in results_AR.predict(0,37)]
     #for i in range(29):
       # if pred[i]<0:
         #   pred[i]=0
   #print(results_AR.fittedvalues)
   #print(pred)
#print(pred[0])
#fig, ax = plt.subplots(figsize=(15,6))
#ax.plot(pred)
   #print(results_AR.predict(30,36))
             #f_pred=results_AR.predict(30,36)
             for name in pro(l):
               if name>=0.09:
                   f_name.append(str(key))
                   break
            
finally:
    with open("./f_name.txt","a+") as f:
        for n in f_name:
            f.write(str(n)+"\n")
    with open("./loss_u.txt","a+") as f_u:
        for u in loss_u:
            f.write(str(u)+"\n")
#print(data_g(data_ac,14808,ac))

speed=====16
speed=====30
speed=====98
speed=====105
speed=====176
wait for .....
wait for .....
wait for .....
wait for .....
wait for .....
speed=====211
speed=====218
speed=====225