Python鏃堕棿搴忓垪鏁版嵁鍒嗘瀽--浠ョず渚嬭鏄?/h1>

Posted 2021-05-02 鏈哄櫒瀛︿範AI绠楁硶宸ョ▼

相关的知识，希望对你有一定的参考价值。

鍚慉I杞瀷鐨勭▼搴忓憳閮藉叧娉ㄤ簡杩欎釜鍙?/span>馃憞馃憞馃憞

鏈枃鐨勫唴瀹逛富瑕佹潵婧愪簬鍗氬锛?br>https://www.analyticsvidhya.com/blog/2016/02/time-series-forecasting-codes-python/ 鑻辨枃涓嶉敊鐨勮鑰呭彲浠ュ墠鍘婚槄璇诲師鏂囥€?/p>

鍦ㄩ槄璇绘湰鏂囦箣鍓?锛屾帹鑽愬厛闃呰锛?/h2>

鏃堕棿搴忓垪棰勬祴涔?-ARIMA妯″瀷

http://www.cnblogs.com/bradleon/p/6827109.html

鏈枃涓昏鍒嗕负鍥涗釜閮ㄥ垎锛?/p>

1. 鐢╬andas澶勭悊鏃跺簭鏁版嵁

2. 鎬庢牱妫€鏌ユ椂搴忔暟鎹殑绋冲畾鎬?/p>

3. 鎬庢牱璁╂椂搴忔暟鎹叿鏈夌ǔ瀹氭€?/p>

4. 鏃跺簭鏁版嵁鐨勯娴?/p>

1. 鐢╬andas瀵煎叆鍜屽鐞嗘椂搴忔暟鎹?/h3>

绗竴姝ワ細瀵煎叆甯哥敤鐨勫簱

import pandas as pd
import numpy as np
import matplotlib.pylab as plt
from matplotlib.pylab 
import rcParams
#rcParams璁惧畾濂界敾甯冪殑澶у皬

rcParams['figure.figsize'] = 15, 6

绗簩姝ワ細瀵煎叆鏃跺簭鏁版嵁
鏁版嵁鏂囦欢鍙湪github锛?/p>

data = pd.read_csv(path+"AirPassengers.csv")
print data.head()
print '\n Data types:'
print data.dtypes

杩愯缁撴灉濡備笅锛氭暟鎹寘鎷瘡涓湀瀵瑰簲鐨刾assenger鐨勬暟鐩€?br>鍙互鐪嬪埌data宸茬粡鏄竴涓?span>DataFrame锛屽寘鍚袱鍒桵onth鍜?Passengers锛屽叾涓璏onth鐨勭被鍨嬫槸object锛岃€宨ndex鏄?,1,2...

绗笁姝ワ細澶勭悊鏃跺簭鏁版嵁
鎴戜滑闇€瑕佸皢Month鐨勭被鍨嬪彉涓?span>datetime锛屽悓鏃朵綔涓篿ndex銆?/p>

dateparse = lambda dates: pd.datetime.strptime(dates, '%Y-%m')
#---鍏朵腑parse_dates 琛ㄦ槑閫夋嫨鏁版嵁涓殑鍝釜column浣滀负date-time淇℃伅锛?br>
#---index_col 鍛婅瘔pandas浠ュ摢涓猚olumn浣滀负 index

#--- date_parser 浣跨敤涓€涓猣unction(鏈枃鐢╨ambda琛ㄨ揪寮忎唬鏇?锛屼娇涓€涓猻tring杞崲涓轰竴涓猟atetime鍙橀噺

data = pd.read_csv('AirPassengers.csv', parse_dates=['Month'], index_col='Month',date_parser=dateparse)
print data.head()
print data.index

缁撴灉濡備笅锛氬彲浠ョ湅鍒癲ata鐨刬ndex宸茬粡鍙樻垚datetime绫诲瀷鐨凪onth浜嗐€?br>2.鎬庢牱妫€鏌ユ椂搴忔暟鎹殑绋冲畾鎬?Stationarity)

鍥犱负ARIMA妯″瀷瑕佹眰鏁版嵁鏄ǔ瀹氱殑锛屾墍浠ヨ繖涓€姝ヨ嚦鍏抽噸瑕併€?/p>

1. 鍒ゆ柇鏁版嵁鏄ǔ瀹氱殑甯稿熀浜庡浜庢椂闂存槸甯搁噺鐨勫嚑涓粺璁￠噺锛?/h4>

1. 甯搁噺鐨勫潎鍊?/p>

2. 甯搁噺鐨勬柟宸?/p>

3. 涓庢椂闂寸嫭绔嬬殑鑷崗鏂瑰樊

鐢ㄥ浘鍍忚鏄庡涓嬶細

1. 鍧囧€?br>2. python鍒ゆ柇鏃跺簭鏁版嵁绋冲畾鎬?/h4>

   鏈変袱绉嶆柟娉曪細
   1.Rolling statistic-- 鍗虫瘡涓椂闂存鍐呯殑骞冲潎鐨勬暟鎹潎鍊煎拰鏍囧噯宸儏鍐点€?/p>

   1. Dickey-Fuller Test -- 杩欎釜姣旇緝澶嶆潅锛屽ぇ鑷存剰鎬濆氨鏄湪涓€瀹氱疆淇℃按骞充笅锛屽浜庢椂搴忔暟鎹亣璁?Null hypothesis: 闈炵ǔ瀹氥€?br>if 閫氳繃妫€楠屽€?statistic)< 涓寸晫鍊?critical value)锛屽垯鎷掔粷null hypothesis锛屽嵆鏁版嵁鏄ǔ瀹氱殑锛涘弽涔嬪垯鏄潪绋冲畾鐨勩€?/p>

   from statsmodels.tsa.stattools import adfuller
   def test_stationarity(timeseries):
       
       #杩欓噷浠ヤ竴骞翠负涓€涓獥鍙ｏ紝姣忎竴涓椂闂磘鐨勫€肩敱瀹冨墠闈?2涓湀锛堝寘鎷嚜宸憋級鐨勫潎鍊间唬鏇匡紝鏍囧噯宸悓鐞嗐€?/span>
       rolmean = pd.rolling_mean(timeseries,window=12)
       rolstd = pd.rolling_std(timeseries, window=12)    
       #plot rolling statistics:
       fig = plt.figure()
       fig.add_subplot()
       orig = plt.plot(timeseries, color = 'blue',label='Original')
       mean = plt.plot(rolmean , color = 'red',label = 'rolling mean')
       std = plt.plot(rolstd, color = 'black', label= 'Rolling standard deviation')
       
       plt.legend(loc = 'best')
       plt.title('Rolling Mean & Standard Deviation')
       plt.show(block=False)    
       
       #Dickey-Fuller test:
       
       print 'Results of Dickey-Fuller Test:'
       dftest = adfuller(timeseries,autolag = 'AIC')    #dftest鐨勮緭鍑哄墠涓€椤逛緷娆′负妫€娴嬪€硷紝p鍊硷紝婊炲悗鏁帮紝浣跨敤鐨勮娴嬫暟锛屽悇涓疆淇″害涓嬬殑涓寸晫鍊?/span>
       dfoutput = pd.Series(dftest[0:4],index = ['Test Statistic','p-value','#Lags Used','Number of Observations Used'])    
   for key,value in dftest[4].items():
           dfoutput['Critical value (%s)' %key] = value    
       print dfoutput
       
   ts = data['#Passengers']
   test_stationarity(ts)

   
   

   缁撴灉濡備笅锛?br>
   

   3. 璁╂椂搴忔暟鎹彉鎴愮ǔ瀹氱殑鏂规硶

   璁╂暟鎹彉寰椾笉绋冲畾鐨勫師鍥犱富瑕佹湁淇╋細

   1. 瓒嬪娍锛坱rend锛?鏁版嵁闅忕潃鏃堕棿鍙樺寲銆傛瘮濡傝鍗囬珮鎴栬€呴檷浣庛€?/p>

   2. 瀛ｈ妭鎬?seasonality)-鏁版嵁鍦ㄧ壒瀹氱殑鏃堕棿娈靛唴鍙樺姩銆傛瘮濡傝鑺傚亣鏃ワ紝鎴栬€呮椿鍔ㄥ鑷存暟鎹殑寮傚父銆?/p>

   鐢变簬鍘熸暟鎹€煎煙鑼冨洿姣旇緝澶э紝涓轰簡缂╁皬鍊煎煙锛屽悓鏃朵繚鐣欏叾浠栦俊鎭紝甯哥敤鐨勬柟娉曟槸瀵规暟鍖?/strong>锛屽彇log銆?/p>

   ts_log = np.log(ts)

   妫€娴嬪拰鍘婚櫎瓒嬪娍
   閫氬父鏈変笁绉嶆柟娉曪細

   • 鑱氬悎 : 灏嗘椂闂磋酱缂╃煭锛屼互涓€娈垫椂闂村唴鏄熸湡/鏈?骞寸殑鍧囧€间綔涓烘暟鎹€笺€備娇涓嶅悓鏃堕棿娈靛唴鐨勫€煎樊璺濈缉灏忋€?/p>

   • 骞虫粦锛?浠ヤ竴涓粦鍔ㄧ獥鍙ｅ唴鐨勫潎鍊间唬鏇垮師鏉ョ殑鍊硷紝涓轰簡浣垮€间箣闂寸殑宸窛缂╁皬

   • 澶氶」寮忚繃婊わ細鐢ㄤ竴涓洖褰掓ā鍨嬫潵鎷熷悎鐜版湁鏁版嵁锛屼娇寰楁暟鎹洿骞虫粦銆?/p>

   鏈枃涓昏浣跨敤骞虫粦鏂规硶

   Moving Average--绉诲姩骞冲潎

   moving_avg = pd.rolling_mean(ts_log,12)
   plt.plot(ts_log ,color = 'blue')
   plt.plot(moving_avg, color='red')

   鍙互鐪嬪嚭moving_average瑕佹瘮鍘熷€煎钩婊戣澶氥€?/p>

   鐒跺悗浣滃樊锛?/p>

   ts_log_moving_avg_diff = ts_log-moving_avg
   ts_log_moving_avg_diff.dropna(inplace = True)
   test_stationarity(ts_log_moving_avg_diff)

   涓婇潰鐨勬柟娉曟槸灏嗘墍鏈夌殑鏃堕棿骞崇瓑鐪嬪緟锛岃€屽湪璁稿鎯呭喌涓嬶紝鍙互璁や负瓒婅繎鐨勬椂鍒昏秺閲嶈銆傛墍浠ュ紩鍏ユ寚鏁板姞鏉冪Щ鍔ㄥ钩鍧?- Exponentially-weighted moving average.锛坧andas涓€氳繃ewma()鍑芥暟鎻愪緵浜嗘鍔熻兘銆傦級

   # halflife鐨勫€煎喅瀹氫簡琛板噺鍥犲瓙alpha锛? alpha = 1 - exp(log(0.5) / halflife)expweighted_avg = pd.ewma(ts_log,halflife=12)
   ts_log_ewma_diff = ts_log - expweighted_avg
   test_stationarity(ts_log_ewma_diff)

   

2. 妫€娴嬪拰鍘婚櫎瀛ｈ妭鎬?br>鏈変袱绉嶆柟娉曪細

• 1 宸垎鍖栵細 浠ョ壒瀹氭粸鍚庢暟鐩殑鏃跺埢鐨勫€肩殑浣滃樊

• 2 鍒嗚В锛?瀵硅秼鍔垮拰瀛ｈ妭鎬у垎鍒缓妯″湪绉婚櫎瀹冧滑

Differencing--宸垎

ts_log_diff = ts_log - ts_log.shift()
ts_log_diff.dropna(inplace=True)
test_stationarity(ts_log_diff)

3.Decomposing-鍒嗚В

#鍒嗚В(decomposing) 鍙互鐢ㄦ潵鎶婃椂搴忔暟鎹腑鐨勮秼鍔垮拰鍛ㄦ湡鎬ф暟鎹兘鍒嗙鍑烘潵:

from statsmodels.tsa.seasonal import seasonal_decompose
def decompose(timeseries):
    
    # 杩斿洖鍖呭惈涓変釜閮ㄥ垎 trend锛堣秼鍔块儴鍒嗭級 锛?seasonal锛堝鑺傛€ч儴鍒嗭級 鍜宺esidual (娈嬬暀閮ㄥ垎)
    decomposition = seasonal_decompose(timeseries)
    
    trend = decomposition.trend
    seasonal = decomposition.seasonal
    residual = decomposition.resid
    
    plt.subplot(411)
    plt.plot(ts_log, label='Original')
    plt.legend(loc='best')
    plt.subplot(412)
    plt.plot(trend, label='Trend')
    plt.legend(loc='best')
    plt.subplot(413)
    plt.plot(seasonal,label='Seasonality')
    plt.legend(loc='best')
    plt.subplot(414)
    plt.plot(residual, label='Residuals')
    plt.legend(loc='best')
    plt.tight_layout()    
    return trend , seasonal, residual

#娑堥櫎浜唗rend 鍜宻easonal涔嬪悗锛屽彧瀵箁esidual閮ㄥ垎浣滀负鎯宠鐨勬椂搴忔暟鎹繘琛屽鐞?/span>trend , seasonal, residual = decompose(ts_log) residual.dropna(inplace=True) test_stationarity(residual)

4. 瀵规椂搴忔暟鎹繘琛岄娴?/h3>

鍋囪缁忚繃澶勭悊锛屽凡缁忓緱鍒颁簡绋冲畾鏃跺簭鏁版嵁銆傛帴涓嬫潵锛屾垜浠娇鐢ˋRIMA妯″瀷
瀵规暟鎹凡缁忛娴嬨€侫RIMA鐨勪粙缁嶅彲浠ヨ鏈洰褰曚笅鐨勫彟涓€绡囨枃绔犮€?/p>

step1锛?閫氳繃ACF,PACF杩涜ARIMA锛坧锛宒锛宷锛夌殑p锛宷鍙傛暟浼拌

鐢卞墠鏂嘍ifferencing閮ㄥ垎宸茬煡锛屼竴闃跺樊鍒嗗悗鏁版嵁宸茬粡绋冲畾锛屾墍浠=1銆?br>鎵€浠ョ敤涓€闃跺樊鍒嗗寲鐨則s_log_diff = ts_log - ts_log.shift() 浣滀负杈撳叆銆?br>绛変环浜?span class="">

yt=Yt鈭?/span>Yt鈭?/span>1

浣滀负杈撳叆銆?/p>

鍏堢敾鍑篈CF,PACF鐨勫浘鍍?浠ｇ爜濡備笅锛?/p>

#ACF and PACF plots:from statsmodels.tsa.stattools import acf, pacf
lag_acf = acf(ts_log_diff, nlags=20)
lag_pacf = pacf(ts_log_diff, nlags=20, method='ols')#Plot ACF: plt.subplot(121) 
plt.plot(lag_acf)
plt.axhline(y=0,linestyle='--',color='gray')
plt.axhline(y=-1.96/np.sqrt(len(ts_log_diff)),linestyle='--',color='gray')
plt.axhline(y=1.96/np.sqrt(len(ts_log_diff)),linestyle='--',color='gray')
plt.title('Autocorrelation Function')#Plot PACF:plt.subplot(122)
plt.plot(lag_pacf)
plt.axhline(y=0,linestyle='--',color='gray')
plt.axhline(y=-1.96/np.sqrt(len(ts_log_diff)),linestyle='--',color='gray')
plt.axhline(y=1.96/np.sqrt(len(ts_log_diff)),linestyle='--',color='gray')
plt.title('Partial Autocorrelation Function')
plt.tight_layout()

鍥句腑锛屼笂涓嬩袱鏉＄伆绾夸箣闂存槸缃俊鍖洪棿锛宲鐨勫€煎氨鏄疉CF绗竴娆＄┛杩囦笂缃俊鍖洪棿鏃剁殑妯酱鍊笺€俼鐨勫€煎氨鏄疨ACF绗竴娆＄┛杩囦笂缃俊鍖洪棿鐨勬í杞村€笺€傛墍浠ヤ粠鍥句腑鍙互寰楀埌p=2锛宷=2銆?/p>

step2锛?寰楀埌鍙傛暟浼拌鍊紁锛宒锛宷涔嬪悗锛岀敓鎴愭ā鍨婣RIMA锛坧锛宒锛宷锛?br>涓轰簡绐佸嚭宸埆锛岀敤涓夌鍙傛暟鍙栧€肩殑涓変釜妯″瀷浣滀负瀵规瘮銆?br>妯″瀷1锛欰R妯″瀷(ARIMA(2,1,0))

from statsmodels.tsa.arima_model import ARIMA
model = ARIMA(ts_log, order=(2, 1, 0))  
results_AR = model.fit(disp=-1)  
plt.plot(ts_log_diff)
plt.plot(results_AR.fittedvalues, color='red')
plt.title('RSS: %.4f'% sum((results_AR.fittedvalues-ts_log_diff)**2))

鍥句腑锛岃摑绾挎槸杈撳叆鍊硷紝绾㈢嚎鏄ā鍨嬬殑鎷熷悎鍊硷紝RSS鐨勭疮璁″钩鏂硅宸€?/p>

妯″瀷2锛歁A妯″瀷锛圓RIMA锛?,1,2锛夛級

model = ARIMA(ts_log, order=(0, 1, 2))  
results_MA = model.fit(disp=-1)  
plt.plot(ts_log_diff)
plt.plot(results_MA.fittedvalues, color='red')
plt.title('RSS: %.4f'% sum((results_MA.fittedvalues-ts_log_diff)**2))

妯″瀷3锛欰RIMA妯″瀷(ARIMA(2,1,2))

model = ARIMA(ts_log, order=(2, 1, 2))  
results_ARIMA = model.fit(disp=-1)  
plt.plot(ts_log_diff)
plt.plot(results_ARIMA.fittedvalues, color='red')
plt.title('RSS: %.4f'% sum((results_ARIMA.fittedvalues-ts_log_diff)**2))

鐢盧SS锛屽彲鐭ユā鍨?--ARIMA锛?,1,2锛夌殑鎷熷悎搴︽渶濂斤紝鎵€浠ユ垜浠‘瀹氫簡鏈€缁堢殑棰勬祴妯″瀷銆?/p>

step3: 灏嗘ā鍨嬩唬鍏ュ師鏁版嵁杩涜棰勬祴
鍥犱负涓婇潰鐨勬ā鍨嬬殑鎷熷悎鍊兼槸瀵瑰師鏁版嵁杩涜绋冲畾鍖栦箣鍚庣殑杈撳叆鏁版嵁鐨勬嫙鍚堬紝鎵€浠ラ渶瑕佸鎷熷悎鍊艰繘琛岀浉搴斿鐞嗙殑閫嗘搷浣滐紝浣垮緱瀹冨洖鍒颁笌鍘熸暟鎹竴鑷寸殑灏哄害銆?/p>

#ARIMA鎷熷悎鐨勫叾瀹炴槸涓€闃跺樊鍒唗s_log_diff锛宲redictions_ARIMA_diff[i]鏄i涓湀涓巌-1涓湀鐨則s_log鐨勫樊鍊笺€?/span>#鐢变簬宸垎鍖栨湁涓€闃舵粸鍚庯紝鎵€浠ョ涓€涓湀鐨勬暟鎹槸绌虹殑锛?br>
predictions_ARIMA_diff = pd.Series(results_ARIMA.fittedvalues, copy=True)print predictions_ARIMA_diff.head()

#绱姞鐜版湁鐨刣iff锛屽緱鍒版瘡涓€间笌绗竴涓湀鐨勫樊鍒嗭紙鍚宭og搴曠殑鎯呭喌涓嬶級銆?br>
#鍗硃redictions_ARIMA_diff_cumsum[i] 鏄i涓湀涓庣1涓湀鐨則s_log鐨勫樊鍊笺€?br>
predictions_ARIMA_diff_cumsum = predictions_ARIMA_diff.cumsum()
#鍏坱s_log_diff => ts_log=>ts_log => ts 

#鍏堜互ts_log鐨勭涓€涓€间綔涓哄熀鏁帮紝澶嶅埗缁欐墍鏈夊€硷紝鐒跺悗姣忎釜鏃跺埢鐨勫€肩疮鍔犱笌绗竴涓湀瀵瑰簲鐨勫樊鍊?杩欐牱灏辫В鍐充簡锛岀涓€涓湀diff鏁版嵁涓虹┖鐨勯棶棰樹簡)

#鐒跺悗寰楀埌浜唒redictions_ARIMA_log => predictions_ARIMA

predictions_ARIMA_log = pd.Series(ts_log.ix[0], index=ts_log.index)
predictions_ARIMA_log = predictions_ARIMA_log.add(predictions_ARIMA_diff_cumsum,fill_value=0)
predictions_ARIMA = np.exp(predictions_ARIMA_log)
plt.figure()
plt.plot(ts)
plt.plot(predictions_ARIMA)
plt.title('RMSE: %.4f'% np.sqrt(sum((predictions_ARIMA-ts)**2)/len(ts)))

5.鎬荤粨

鍓嶉潰涓€绡囨枃绔狅紝鎬荤粨浜咥RIMA寤烘ā鐨勬楠ゃ€?br>(1). 鑾峰彇琚娴嬬郴缁熸椂闂村簭鍒楁暟鎹紱
(2). 瀵规暟鎹粯鍥撅紝瑙傛祴鏄惁涓哄钩绋虫椂闂村簭鍒楋紱瀵逛簬闈炲钩绋虫椂闂村簭鍒楄鍏堣繘琛宒闃跺樊鍒嗚繍绠楋紝鍖栦负骞崇ǔ鏃堕棿搴忓垪锛?br>(3). 缁忚繃绗簩姝ュ鐞嗭紝宸茬粡寰楀埌骞崇ǔ鏃堕棿搴忓垪銆傝瀵瑰钩绋虫椂闂村簭鍒楀垎鍒眰寰楀叾鑷浉鍏崇郴鏁癆CF 鍜屽亸鑷浉鍏崇郴鏁癙ACF锛岄€氳繃瀵硅嚜鐩稿叧鍥惧拰鍋忚嚜鐩稿叧鍥剧殑鍒嗘瀽锛屽緱鍒版渶浣崇殑闃跺眰 p 鍜岄樁鏁?q
(4). 鐢变互涓婂緱鍒扮殑d銆乹銆乸锛屽緱鍒癆RIMA妯″瀷銆傜劧鍚庡紑濮嬪寰楀埌鐨勬ā鍨嬭繘琛屾ā鍨嬫楠屻€?br>鍏蜂綋渚嬪瓙浼氬湪鍙︿竴绡囨枃绔犱腑缁欏嚭銆?/p>

鏈枃缁撳悎涓€涓緥瀛愶紝璇存槑python濡備綍瑙ｅ喅锛?br>1.鍒ゆ柇涓€涓椂搴忔暟鎹槸鍚︽槸绋冲畾銆傚搴旀楠?1)

1. 鎬庢牱璁╂椂搴忔暟鎹ǔ瀹氬寲銆傚搴旀楠?2)

2. 浣跨敤ARIMA妯″瀷杩涜鏃跺簭鏁版嵁棰勬祴銆傚搴旀楠?3,4)

鍙﹀瀵筪ata science鎰熷叴瓒ｇ殑鍚屽鍙互鍏虫敞杩欎釜缃戠珯锛屽共璐ц繕鎸哄鐨勩€?br>https://www.analyticsvidhya.com/blog/

娣卞害瀛︿範銆佹満鍣ㄥ涔犮€佹暟鎹垎鏋愩€乸ython

以上是关于Python鏃堕棿搴忓垪鏁版嵁鍒嗘瀽--浠ョず渚嬭鏄?/h1>

的主要内容，如果未能解决你的问题，请参考以下文章

鍟嗕笟鏁版嵁鍒嗘瀽鍜屽彲瑙嗗寲BI锛岄浂鍩虹鍏ラ棬闇€瑕佸涔咃紵

鏁版嵁鍒嗘瀽濡備綍鍔╁姏淇濋櫓琛屼笟瀹炵幇鐪熸鐨勨€滄寜闇€鈥濇湇鍔★紵

寤虹珛浣犳暟鎹垎鏋愮殑鎬濈淮妗嗘灦

鏁版嵁搴忓垪鍖栨牸寮?YAML

Echars璁茶В鍒嗘瀽