added execution pipeline

alphamind/execution/pipeline.py

@@ -21,7 +21,7 @@ class ExecutionPipeline(object):
         turn_over, planed_pos = 0., target_pos
 
         for executor in self.executors:
-            turn_over, planed_pos = executor.execute(target_pos)
+            turn_over, planed_pos = executor.execute(planed_pos)
 
         executed_pos = planed_pos
 

alphamind/execution/targetvolexecutor.py

@@ -8,6 +8,7 @@ Created on 2017-9-22
 from typing import Tuple
 import pandas as pd
 from PyFin.Math.Accumulators import MovingStandardDeviation
+from PyFin.Math.Accumulators import MovingAverage
 from alphamind.execution.baseexecutor import ExecutorBase
 
 
@@ -16,6 +17,7 @@ class TargetVolExecutor(ExecutorBase):
     def __init__(self, window=30, target_vol=0.01):
         super().__init__()
         self.m_vol = MovingStandardDeviation(window=window, dependency='return')
+        self.m_leverage = MovingAverage(window=window, dependency='leverage')
         self.target_vol = target_vol
         self.multiplier = 1.
 
@@ -28,11 +30,16 @@ class TargetVolExecutor(ExecutorBase):
             return turn_over, target_pos
         else:
             c_vol = self.m_vol.result()
-            self.multiplier = c_vol / self.target_vol
+            c_leverage = self.m_leverage.result()
+            self.multiplier = self.target_vol / c_vol * c_leverage
             candidate_pos = target_pos.copy()
-            candidate_pos['weight'] = candidate_pos.weight.values / self.multiplier
+            candidate_pos['weight'] = candidate_pos.weight.values * self.multiplier
             turn_over = self.calc_turn_over(candidate_pos, self.current_pos)
             return turn_over, candidate_pos
 
+    def set_current(self, current_pos: pd.DataFrame):
+        super().set_current(current_pos)
+        self.m_leverage.push({'leverage': current_pos.weight.abs().sum()})
+
     def update(self, data_dict: dict):
         self.m_vol.push(data_dict)

alphamind/tests/execution/test_pipeline.py

+# -*- coding: utf-8 -*-
+"""
+Created on 2017-9-25
+
+@author: cheng.li
+"""
+
+import unittest
+from collections import deque
+import numpy as np
+import pandas as pd
+from alphamind.execution.pipeline import ExecutionPipeline
+from alphamind.execution.thresholdexecutor import ThresholdExecutor
+from alphamind.execution.targetvolexecutor import TargetVolExecutor
+
+
+class TestExecutionPipelin(unittest.TestCase):
+
+    def test_execution_pipeline(self):
+        n = 100
+        window = 60
+        target_vol = 0.01
+        turn_over_threshold = 0.5
+
+        executor1 = TargetVolExecutor(window=window, target_vol=target_vol)
+        executor2 = ThresholdExecutor(turn_over_threshold=turn_over_threshold)
+
+        execution_pipeline = ExecutionPipeline(executors=[executor1, executor2])
+
+        return_1 = np.random.randn(2000, n) * 0.05
+        return_2 = np.random.randn(2000, n) * 0.2
+        return_total = np.concatenate((return_1, return_2))
+        codes = np.array(list(range(n)))
+
+        ret_deq = deque(maxlen=window)
+
+        for i, row in enumerate(return_total):
+            weights = np.random.randint(0, 100, n)
+            weights = weights / weights.sum()
+            pos = pd.DataFrame({'code': codes, 'weight': weights})
+            turn_over, executed_pos = execution_pipeline.execute(pos)
+            daily_return = row @ executed_pos.weight.values.flatten()
+            data_dict = {'return': daily_return}
+            execution_pipeline.update(data_dict=data_dict)
+            ret_deq.append(daily_return)

alphamind/tests/execution/test_targetvolexecutor.py

@@ -25,13 +25,13 @@ class TestTargetVolExecutor(unittest.TestCase):
         return_2 = np.random.randn(2000, n) * 0.2
         return_total = np.concatenate((return_1, return_2))
 
-        weigths = np.ones(n) / n
+        weights = np.ones(n) / n
         codes = np.array(list(range(n)))
 
         ret_deq = deque(maxlen=window)
 
         for i, row in enumerate(return_total):
-            pos = pd.DataFrame({'code': codes, 'weight': weigths})
+            pos = pd.DataFrame({'code': codes, 'weight': weights})
             turn_over, executed_pos = executor.execute(pos)
 
             if i >= window:
@@ -41,7 +41,7 @@ class TestTargetVolExecutor(unittest.TestCase):
                 executed_pos.equals(pos)
 
             executor.set_current(executed_pos)
-            daily_return = row @ weigths
+            daily_return = row @ weights
             data_dict = {'return': daily_return}
             executor.update(data_dict=data_dict)
             ret_deq.append(daily_return)