- Refactoring of ARIMA façade for statsmodels

- QuantReg façade for statsmodels
 - EnsembleFTS

benchmarks/arima.py

@@ -3,6 +3,7 @@
 
 import numpy as np
 from statsmodels.tsa.arima_model import ARIMA as stats_arima
+from statsmodels.tsa.arima_model import ARMA
 from pyFTS import fts
 
 
@@ -31,24 +32,43 @@ class ARIMA(fts.FTS):
 
         old_fit = self.model_fit
         self.model =  stats_arima(data, order=(self.p, self.d, self.q))
-        try:
+        #try:
-            self.model_fit = self.model.fit(disp=0)
+        self.model_fit = self.model.fit(disp=0)
-        except:
+        #except:
-            try:
+        #    try:
-                self.model = stats_arima(data, order=(self.p, self.d, self.q))
+        #        self.model = stats_arima(data, order=(self.p, self.d, self.q))
-                self.model_fit = self.model.fit(disp=1)
+        #        self.model_fit = self.model.fit(disp=1)
-            except:
+        #    except:
-                self.model_fit = old_fit
+        #        self.model_fit = old_fit
 
-        self.trained_data = data #.tolist()
+        #self.trained_data = data #.tolist()
+
+    def ar(self, data):
+        return data.dot(self.model_fit.arparams)
+
+    def ma(self, data):
+        return data.dot(self.model_fit.maparams)
 
     def forecast(self, data):
         if self.model_fit is None:
             return np.nan
+
+        order = self.p
+
+        ndata = np.array(self.doTransformations(data))
+
+        l = len(ndata)
+
         ret = []
-        for t in data:
+
-            output = self.model_fit.forecast()
+        ar = np.array([self.ar(ndata[k - self.p: k]) for k in np.arange(self.p, l)])
-            ret.append( output[0] )
+
-            self.trained_data = np.append(self.trained_data, t) #.append(t)
+        residuals = np.array([ar[k - self.p] - ndata[k] for k in np.arange(self.p, l)])
-            self.train(self.trained_data,None,order=self.order, parameters=(self.p, self.d, self.q))
+
-        return ret
+        ma = np.array([self.ma(residuals[k - self.q : k]) for k in np.arange(self.q, len(ar)+1)])
+
+        ret = ar + ma
+
+        ret = self.doInverseTransformations(ret, params=[data[order - 1:]])
+
+        return ret

benchmarks/quantreg.py

@@ -0,0 +1,24 @@
+#!/usr/bin/python
+# -*- coding: utf8 -*-
+
+import numpy as np
+from statsmodels.regression.quantile_regression import QuantReg
+from pyFTS import fts
+
+
+class QuantileRegression(fts.FTS):
+    def __init__(self, name):
+        super(QuantileRegression, self).__init__(1, "QR")
+        self.name = "QR"
+        self.detail = "Quantile Regression"
+        self.isHighOrder = True
+        self.hasIntervalForecasting = True
+        self.benchmark_only = True
+        self.minOrder = 1
+        self.alpha = 0.5
+
+    def train(self, data, sets, order=1, parameters=None):
+        pass
+
+    def forecast(self, data):
+        pass

ensemble.py

@@ -0,0 +1,93 @@
+#!/usr/bin/python
+# -*- coding: utf8 -*-
+
+import numpy as np
+import pandas as pd
+import math
+from operator import itemgetter
+from pyFTS.common import FLR, FuzzySet, SortedCollection
+from pyFTS import fts
+
+class EnsembleFTS(fts.FTS):
+    def __init__(self, name, update=True):
+        super(EnsembleFTS, self).__init__("Ensemble FTS")
+        self.shortname = "Ensemble FTS " + name
+        self.name = "Ensemble FTS"
+        self.flrgs = {}
+        self.hasPointForecasting = True
+        self.hasIntervalForecasting = True
+        self.hasDistributionForecasting = True
+        self.isHighOrder = True
+        self.models = []
+        self.parameters = []
+
+    def train(self, data, sets, order=1,parameters=None):
+
+        pass
+
+    def forecast(self, data):
+
+        ndata = np.array(self.doTransformations(data))
+
+        l = len(ndata)
+
+        ret = []
+
+        for k in np.arange(self.order - 1, l):
+            pass
+
+        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):
+            pass
+
+        return ret
+
+    def forecastAhead(self, data, steps):
+        pass
+
+    def forecastAheadInterval(self, data, steps):
+        pass
+
+
+    def getGridClean(self, resolution):
+        grid = {}
+
+        if len(self.transformations) == 0:
+            _min = self.sets[0].lower
+            _max = self.sets[-1].upper
+        else:
+            _min = self.original_min
+            _max = self.original_max
+
+        for sbin in np.arange(_min,_max, resolution):
+            grid[sbin] = 0
+
+        return grid
+
+    def gridCount(self, grid, resolution, index, interval):
+        #print(interval)
+        for k in index.inside(interval[0],interval[1]):
+            #print(k)
+            grid[k] += 1
+        return grid
+
+    def gridCountPoint(self, grid, resolution, index, point):
+        k = index.find_ge(point)
+        # print(k)
+        grid[k] += 1
+        return grid
+
+    def forecastAheadDistribution(self, data, steps, resolution, parameters=2):
+        pass
+

tests/general.py

@@ -28,18 +28,46 @@ os.chdir("/home/petronio/dados/Dropbox/Doutorado/Codigos/")
 taiexpd = pd.read_csv("DataSets/TAIEX.csv", sep=",")
 taiex = np.array(taiexpd["avg"][:5000])
 
-from pyFTS.benchmarks import distributed_benchmarks as bchmk
+from statsmodels.tsa.arima_model import ARIMA as stats_arima
+
+model = stats_arima(taiex[:1600], (2,0,1)).fit(disp=0)
+
+ar = np.array(taiex[1598:1600]).dot( model.arparams )
+
+#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])
+
+
+#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 import yu
+from pyFTS.benchmarks import arima
+
+
+tmp = arima.ARIMA("")
+tmp.train(taiex[:1600],None,parameters=(2,0,1))
+teste = tmp.forecast(taiex[1598:1601])
+
+print(teste)
 
 #bchmk.teste(taiex,['192.168.0.109', '192.168.0.101'])
 
-bchmk.point_sliding_window(taiex,2000,train=0.8, #models=[yu.WeightedFTS], # #
+#bchmk.point_sliding_window(taiex,2000,train=0.8, #models=[yu.WeightedFTS], # #
-                     partitioners=[Grid.GridPartitioner], #Entropy.EntropyPartitioner], # FCM.FCMPartitioner, ],
+#                     partitioners=[Grid.GridPartitioner], #Entropy.EntropyPartitioner], # FCM.FCMPartitioner, ],
-                     partitions= np.arange(10,200,step=5), #transformation=diff,
+#                     partitions= np.arange(10,200,step=5), #transformation=diff,
-                     dump=False, save=True, file="experiments/nasdaq_point_distributed.csv",
+#                     dump=False, save=True, file="experiments/nasdaq_point_distributed.csv",
-                     nodes=['192.168.0.109', '192.168.0.101']) #, depends=[hofts, ifts])
+#                     nodes=['192.168.0.109', '192.168.0.101']) #, depends=[hofts, ifts])
 
 #bchmk.testa(taiex,[10,20],partitioners=[Grid.GridPartitioner], nodes=['192.168.0.109', '192.168.0.101'])