Detecting Invalid Associations between Fare Machines and Metro Stations Using Smart Card Data

Table 1

Model parameters.

Tensor decomposition	Optimal value (potential values)
Number of components (R)	8 (1–15)
Optimization algorithm	(ALS refers to the alternating least squares algorithm)
Error tolerance	1e − 6 (1e − 3–1e − 8)
Maximum number of iterations	100 (10, 100, 500, 1000)
Isolation forest	Value
Threshold score (the threshold score is calculated with the decision_function in sklearn.IsolationForest package under Python 3.7)	0
Number of estimators	1000 (200, 500, 1000, 1500)
Station-NN	Value
The number of top stations in P1 reinspection	5
Number of hidden layers	2 (1–5)
Optimizer	Adam (Adam refers to the optimization algorithm proposed in [29])
Number of neurons	(16, 5)