To install "sudo pip3 install git+https://github.com/captify/bonspy"
Bonspy converts bidding trees from various input formats to the Bonsai bidding language of AppNexus.
As intermediate format bonspy constructs a NetworkX graph from which it produces the Bonsai language output. Bidding trees may also be constructed directly in this NetworkX format (see first example below).
At present bonspy provides a converter from trained sklearn logistic regression classifiers with categorical, one-hot encoded features to the intermediate NetworkX format (see second example below).
In combination with our AppNexus API wrapper nexusadspy it is also
straightforward to check your bidding tree for syntactical errors and upload it for real-time bidding (third example below).
This package was developed and tested on Python 3.5. However, the examples below have been tested successfully in Python 2.7.
import networkx as nx
from bonspy import BonsaiTree
import numpy as np
def generateBucketAPBtree(segmentID, bids, header=''):
g2 = nx.DiGraph()
nLeafs = len(bids)
g2.add_node(0, split='segment.value', state={'segment': segmentID})
for i in range(nLeafs):
g2.add_node(i + 1, is_leaf=True, is_smart=True, leaf_name=str(nLeafs - i), value=float(bids[i]),
state={'segment': segmentID})
g2.add_edge(0, i + 1, value=(nLeafs - i, None), type='range')
g2.add_node(nLeafs + 1, is_leaf=True, is_smart=True, is_default_leaf=True, value=0, state={'segment': segmentID})
g2.add_edge(0, nLeafs + 1, type='assignment')
tree2 = BonsaiTree(g2)
theTreeWithHeader = header + tree2.bonsai
return theTreeWithHeader
segmentID = 199
bids = np.array([0, 0, 0, 1, 2, 3, 4, 0, 0])
header = '''#Gererated by some optimiser
#rna: no')'''
theTree = generateBucketAPBtree(segmentID, bids, header)
print(theTree)
Note that non-leaf nodes track the next user variable to be split on in their split attribute while
the current choice of user features is tracked in their state attribute.
Leaves designate their output (the bid) in their output attribute.
The Bonsai text representation of the above tree is stored in its .bonsai attribute:
print(tree.bonsai)
prints out
#Gererated by some optimiser
#rna: no')
if segment[199].value >= 9:
leaf_name: "9"
value: no_bid
elif segment[199].value >= 8:
leaf_name: "8"
value: no_bid
elif segment[199].value >= 7:
leaf_name: "7"
value: no_bid
elif segment[199].value >= 6:
leaf_name: "6"
value: 1.0000
elif segment[199].value >= 5:
leaf_name: "5"
value: 2.0000
elif segment[199].value >= 4:
leaf_name: "4"
value: 3.0000
elif segment[199].value >= 3:
leaf_name: "3"
value: 4.0000
elif segment[199].value >= 2:
leaf_name: "2"
value: no_bid
elif segment[199].value >= 1:
leaf_name: "1"
value: no_bid
else:
value: no_bid
nodeMap = {}
def getNodeID(C, A, R):
key = str(C) + str(A) + str(R)
if key not in nodeMap.keys():
node = len(nodeMap)
nodeMap.update({key: node})
return nodeMap[key]
def generateSR_APBtree(segmentIDs, ages, header=''):
g2 = nx.DiGraph()
nSegments = len(segmentIDs)
nAges = len(ages)
g2.add_node(getNodeID(0, 0, 0), split='segment', state={})
for iSeg in range(nSegments):
g2.add_node(getNodeID(iSeg + 1, 0, 0), split='segment.age', state={'segment': segmentIDs[iSeg]})
g2.add_edge(getNodeID(0, 0, 0), getNodeID(iSeg + 1, 0, 0), value=segmentIDs[iSeg], type='assignment')
for iAge in range(nAges):
g2.add_node(getNodeID(iSeg + 1, iAge + 1, 0), is_leaf=True, is_smart=True, leaf_name=str(iAge), value=iAge,
state={'segment': segmentIDs[iSeg], 'segment.age': ages[iAge]})
g2.add_edge(getNodeID(iSeg + 1, 0, 0), getNodeID(iSeg + 1, iAge + 1, 0), value=(iAge, None), type='range',
join_statement=True, is_negated=True)
g2.add_node(getNodeID(iSeg + 1, nAges, 0), is_leaf=True, is_smart=True, is_default_leaf=True, value=1,
state={'segment': segmentID})
g2.add_edge(getNodeID(iSeg + 1, 0, 0), getNodeID(iSeg + 1, nAges, 0), type='assignment')
tree2 = BonsaiTree(g2)
theTreeWithHeader = header + tree2.bonsai
return theTreeWithHeader
features = {
'age': [10, 15, 30, 60, 180],
'recency': [10, 30, 60],
'geo': ['FR, UK, UA']
}
segmentIDs = [10898030, 10898031, 10898032]
ages = [10, 15, 30, 60, 180]
header = '''#Gererated by some optimiser
#rna: no\n'''
theTree = generateSR_APBtree(segmentIDs, ages, header=header)
print(theTree)
Prints out
#Gererated by some optimiser
#rna: no
if segment[10898030]:
if not every segment[10898030].age >= 0:
leaf_name: "0"
value: no_bid
elif not every segment[10898030].age >= 1:
leaf_name: "1"
value: 1.0000
elif not every segment[10898030].age >= 2:
leaf_name: "2"
value: 2.0000
elif not every segment[10898030].age >= 3:
leaf_name: "3"
value: 3.0000
else:
leaf_name: "4"
value: 1.0000
elif segment[10898031]:
if not every segment[10898031].age >= 0:
leaf_name: "0"
value: no_bid
elif not every segment[10898031].age >= 1:
leaf_name: "1"
value: 1.0000
elif not every segment[10898031].age >= 2:
leaf_name: "2"
value: 2.0000
elif not every segment[10898031].age >= 3:
leaf_name: "3"
value: 3.0000
else:
leaf_name: "4"
value: 1.0000
else segment[10898032]:
if not every segment[10898032].age >= 0:
leaf_name: "0"
value: no_bid
elif not every segment[10898032].age >= 1:
leaf_name: "1"
value: 1.0000
elif not every segment[10898032].age >= 2:
leaf_name: "2"
value: 2.0000
elif not every segment[10898032].age >= 3:
leaf_name: "3"
value: 3.0000
else:
leaf_name: "4"
value: 1.0000
For trees with arbitraty number of features:
def test():
from bonspy import LogisticConverter
from bonspy import BonsaiTree
features = ['segment', 'segment.age', 'campaign', 'campaign.recency']
vocabulary = {
'segment=10898030': 0,
'segment=10898031': 1,
'segment.age=0': 2,
'segment.age=1': 3,
'segment.age=2': 4,
'segment.age=3': 5,
'campaign=25198998': 6,
'campaign.recency=10': 7,
'campaign.recency=20': 8
}
weights = [.1, .2, .15, .25, .1, .1, .2, .2, .2, .2]
intercept = .4
buckets = {
'segment.age': {
'0': (None, 10),
'1': (None, 15),
'2': (None, 30),
'3': (None, 60)
}
}
types = {
'segment': 'assignment',
'segment.age': 'range',
'campaign': 'assignment',
'campaign.recency': 'assignment'
}
conv = LogisticConverter(features=features, vocabulary=vocabulary,
weights=weights, intercept=intercept,
types=types, base_bid=2., buckets=buckets)
tree = BonsaiTree(conv.graph)
print(tree.bonsai)
test()
Prints out
if segment[10898030]:
if segment[10898030].age <= 10:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.4011
else:
1.3140
elif segment[10898030].age <= 15:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.4422
else:
1.3584
elif segment[10898030].age <= 30:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.3799
else:
1.2913
elif segment[10898030].age <= 60:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.3799
else:
1.2913
else:
1.2449
elif segment[10898031]:
if segment[10898031].age <= 10:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.4422
else:
1.3584
elif segment[10898031].age <= 15:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.4815
else:
1.4011
elif segment[10898031].age <= 30:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.4219
else:
1.3364
elif segment[10898031].age <= 60:
if campaign[25198998]:
if campaign[25198998].recency=10:
leaf_name: "blah"
value: 2.0000
elif campaign[25198998].recency=20:
leaf_name: "blah"
value: 2.0000
else:
1.4219
else:
1.3364
else:
1.2913
else:
1.1974