- QuantReg façade for statsmodels

benchmarks/arima.py

@@ -23,6 +23,8 @@ class ARIMA(fts.FTS):
         self.minOrder = 1
 
     def train(self, data, sets, order=1, parameters=None):
+        ndata = np.array(self.doTransformations(data))
+
         if parameters is not None:
             self.p = parameters[0]
             self.d = parameters[1]
@@ -31,7 +33,7 @@ class ARIMA(fts.FTS):
             self.shortname = "ARIMA(" + str(self.p) + "," + str(self.d) + "," + str(self.q) + ")"
 
         old_fit = self.model_fit
-        self.model =  stats_arima(data, order=(self.p, self.d, self.q))
+        self.model =  stats_arima(ndata, order=(self.p, self.d, self.q))
         #try:
         self.model_fit = self.model.fit(disp=0)
         #except:

benchmarks/quantreg.py

@@ -3,6 +3,7 @@
 
 import numpy as np
 from statsmodels.regression.quantile_regression import QuantReg
+from statsmodels.tsa.tsatools import lagmat
 from pyFTS import fts
 
 
@@ -12,13 +13,60 @@ class QuantileRegression(fts.FTS):
         self.name = "QR"
         self.detail = "Quantile Regression"
         self.isHighOrder = True
+        self.hasPointForecasting = True
         self.hasIntervalForecasting = True
         self.benchmark_only = True
         self.minOrder = 1
         self.alpha = 0.5
+        self.upper_qt = None
+        self.mean_qt = None
+        self.lower_qt = None
 
     def train(self, data, sets, order=1, parameters=None):
-        pass
+        self.order = order
+        tmp = np.array(self.doTransformations(data))
+
+        lagdata, ndata = lagmat(tmp, maxlag=order, trim="both", original='sep')
+        uqt = QuantReg(ndata, lagdata).fit(1 - self.alpha)
+        mqt = QuantReg(ndata, lagdata).fit(0.5)
+        lqt = QuantReg(ndata, lagdata).fit(self.alpha)
+
+        self.upper_qt = [uqt.params[k] for k in uqt.params.keys()]
+        self.mean_qt = [mqt.params[k] for k in mqt.params.keys()]
+        self.lower_qt = [lqt.params[k] for k in lqt.params.keys()]
+
+    def linearmodel(self,data,params):
+        return params[0] + sum([ data[k] * params[k+1] for k in np.arange(0, self.order) ])
 
     def forecast(self, data):
-        pass
+        ndata = np.array(self.doTransformations(data))
+        l = len(ndata)
+
+        ret = []
+
+        for k in np.arange(self.order, l):
+            sample = ndata[k - self.order : k]
+
+            ret.append(self.linearmodel(sample, self.mean_qt))
+
+        ret = self.doInverseTransformations(ret, params=[data[self.order - 1:]])
+
+        return ret
+
+    def forecastInterval(self, data):
+
+        ndata = np.array(self.doTransformations(data))
+
+        l = len(ndata)
+
+        ret = []
+
+        for k in np.arange(self.order - 1, l):
+            sample = ndata[k - self.order: k]
+            up = self.linearmodel(sample, self.upper_qt)
+            down = self.linearmodel(sample, self.down_qt)
+            ret.append([up, down])
+
+        ret = self.doInverseTransformations(ret, params=[data[self.order - 1:]])
+
+        return ret

tests/general.py

@@ -28,38 +28,27 @@ os.chdir("/home/petronio/dados/Dropbox/Doutorado/Codigos/")
 taiexpd = pd.read_csv("DataSets/TAIEX.csv", sep=",")
 taiex = np.array(taiexpd["avg"][:5000])
 
-from statsmodels.tsa.arima_model import ARIMA as stats_arima
+#from statsmodels.tsa.arima_model import ARIMA as stats_arima
+from statsmodels.tsa.tsatools import lagmat
 
-model = stats_arima(taiex[:1600], (2,0,1)).fit(disp=0)
+tmp = np.arange(10)
 
-ar = np.array(taiex[1598:1600]).dot( model.arparams )
+lag, a = lagmat(tmp, maxlag=2, trim="both", original='sep')
-
-#print(ar)
-
-res = ar - taiex[1600]
-
-#print(res)
-
-ma = np.array([res]).dot(model.maparams)
-
-#print(ma)
-
-print(ar + ma)
-print(taiex[1598:1601])
-print(taiex[1600])
 
+print(lag)
+print(a)
 
 #from pyFTS.benchmarks import distributed_benchmarks as bchmk
 #from pyFTS.benchmarks import parallel_benchmarks as bchmk
 #from pyFTS.benchmarks import benchmarks as bchmk
-from pyFTS.benchmarks import arima
+#from pyFTS.benchmarks import arima
 
 
-tmp = arima.ARIMA("")
+#tmp = arima.ARIMA("")
-tmp.train(taiex[:1600],None,parameters=(2,0,1))
+#tmp.train(taiex[:1600],None,parameters=(2,0,1))
-teste = tmp.forecast(taiex[1598:1601])
+#teste = tmp.forecast(taiex[1598:1601])
 
-print(teste)
+#print(teste)
 
 #bchmk.teste(taiex,['192.168.0.109', '192.168.0.101'])