Bagged Tree Model

Ensemble decision tree model built using bagging (bootstrap aggregation) sampling technique on gas sensor array data. Results are presented below.

Misclassification Rate - 1.1% (on test data)

Bagged Tree Confusion Matrix

Bagged Tree Error As a Function of Number of Trees

OOB = Out of Bag Error

Variable Importance

Examples of how to read feature names on the above chart

S2DR - Direct Resistance of sensor 2

S11NDR - Normalized Direct Resistance of sensor 11

S9I_1 - Exponential Moving Average with a smoothing parameter of 0.1 when current is increasing for Sensor 9

S9D_01 - Exponential Moving Average with a smoothing parameter of 0.01 when current is decreasing for Sensor 9

CONC - Concentration (ppmv)

ROC Curve

Area Under Curve

0.9995388