neuralnetnode properties
Important: A newer version of the Neural Net modeling node, with enhanced features, is
available in this release and is described in the next section (neuralnetwork). Although you
can still build and score a model with the previous version, we recommend updating your scripts to
use the new version. Details of the previous version are retained here for reference.
Example
node = stream.create("neuralnet", "My node")
# "Fields" tab
node.setPropertyValue("custom_fields", True)
node.setPropertyValue("targets", ["Drug"])
node.setPropertyValue("inputs", ["Age", "Na", "K", "Cholesterol", "BP"])
# "Model" tab
node.setPropertyValue("use_partitioned_data", True)
node.setPropertyValue("method", "Dynamic")
node.setPropertyValue("train_pct", 30)
node.setPropertyValue("set_random_seed", True)
node.setPropertyValue("random_seed", 12345)
node.setPropertyValue("stop_on", "Time")
node.setPropertyValue("accuracy", 95)
node.setPropertyValue("cycles", 200)
node.setPropertyValue("time", 3)
node.setPropertyValue("optimize", "Speed")
# "Multiple Method Expert Options" section
node.setPropertyValue("m_topologies", "5 30 5; 2 20 3, 1 10 1")
node.setPropertyValue("m_non_pyramids", False)
node.setPropertyValue("m_persistence", 100)
neuralnetnode Properties |
Values | Property description |
|---|---|---|
targets
|
[field1 ... fieldN] | The Neural Net node expects one or more target fields and one or more input fields. Frequency and weight fields are ignored. See the topic Common modeling node properties for more information. |
method
|
Quick
Dynamic
Multiple
Prune
ExhaustivePrune
RBFN
|
|
prevent_overtrain
|
flag | |
train_pct
|
number | |
set_random_seed
|
flag | |
random_seed
|
number | |
mode
|
Simple
Expert
|
|
stop_on
|
Default
Accuracy
Cycles
Time
|
Stopping mode. |
accuracy
|
number | Stopping accuracy. |
cycles
|
number | Cycles to train. |
time
|
number | Time to train (minutes). |
continue
|
flag | |
show_feedback
|
flag | |
binary_encode
|
flag | |
use_last_model
|
flag | |
gen_logfile
|
flag | |
logfile_name
|
string | |
alpha
|
number | |
initial_eta
|
number | |
high_eta
|
number | |
low_eta
|
number | |
eta_decay_cycles
|
number | |
hid_layers
|
One
Two
Three
|
|
hl_units_one
|
number | |
hl_units_two
|
number | |
hl_units_three
|
number | |
persistence
|
number | |
m_topologies
|
string | |
m_non_pyramids
|
flag | |
m_persistence
|
number | |
p_hid_layers
|
One
Two
Three
|
|
p_hl_units_one
|
number | |
p_hl_units_two
|
number | |
p_hl_units_three
|
number | |
p_persistence
|
number | |
p_hid_rate
|
number | |
p_hid_pers
|
number | |
p_inp_rate
|
number | |
p_inp_pers
|
number | |
p_overall_pers
|
number | |
r_persistence
|
number | |
r_num_clusters
|
number | |
r_eta_auto
|
flag | |
r_alpha
|
number | |
r_eta
|
number | |
optimize
|
Speed
Memory
|
Use to specify whether model building should be optimized for speed or for memory. |
calculate_variable_importance
|
flag | Note: The sensitivity_analysis property used in previous releases is
deprecated in favor of this property. The old property is still supported, but
calculate_variable_importance is recommended. |
calculate_raw_propensities
|
flag | |
calculate_adjusted_propensities
|
flag | |
adjusted_propensity_partition
|
Test
Validation
|