TimeSeriesInterpretorRNN 0.0.11

No project description provided

TimeSeriesRNNInterpretor

TimeSeriesRNNInterpretor, Interprets the RNN black box model. It internally uses shap.DeepExplainer. We have tried to interpret only single instance with respect to time. We have additionally included monthly effect of the instance.

• ✨TimeSeriesRNNInterpretor✨

Parameters

Parameters required to Pass

Parameter	To be passed	Format
model	RNN model	tensorflow V1 behaviour
background	training data sample	Numpy array
test	Instance data which we want to explain	Numpy array
features	feature_names or column names	df.columns
max_output	max output value(1)	1 for binary classification
max_display	max display features for waterfall plot	a number < no of features
decision_range	max display features for decision plot	a number < no of features
Measure	"absolute" or "relative" or "total "	for plotting waterfall plot w.r.t basevalue
dependence_feature_3	3rd feature name to plot dependence plot	a feature name or index of feature
dependence_feature_2	2nd feature name to plot dependence plot	a feature name or index of the feature

Calling

from TimeSeriesRNNInterpretor import TimeSeriesRNNInterpretor

temp = TimeSeriesInterpretorRNN(model=model,background=background,test=test,features=features,max_output=1,max_display=15,decision_range=30,Measure="absolute",dependence_feature_3="age.",dependence_feature_2="gender")

Plotting Different Plots

temp.global_patient() # To the know the feature importance of a instance
temp.monthly_chart() # timeframe charts of a instance (day,month, or any time frame)
temp.two_month_compare(17, 18) # Specify two timeframe numbers of a instancee to compare (days,months, or any time frames)
temp.single_month(18) # Specify the required specific timeframe number of instance(day,month, or any time frame)