We propose to use the neural network model (NN), which gives a direct estimation of occupancy rates. In this proposal we indicate that it is difficult to output the limits 1.0, 0.0 of the Sigmoid function, which has up until now often been used. As a way to improve this, we explain a new method, analogue output NN which uses output layer response function as a linear function. We assessed it using a simulation test and practical data collected prior to this from Ashigara, S.A. and the effect was verified. We were able to find that the result came within 10% of the percent occupancy rate's standard deviation from the true value and were able to obtain sufficient possibilities for practical use. Based on this positive result, this method was adopted for Niikura, P.A. and moreover the effect when practically applied was also verified. As a result of this verification, the evaluation is the same as the practical data in Ashigara, S.A. and we were able to obtain the positive result of the standard deviation from the true value being within 10% of occupancy rates.
A direct estimation method of occupancy rate in parking lot using analogue output neural network model
Ein direktes Schätzverfahren für die Belegung von Parkplätzen mit Hilfe eines neuronalen Netzmodells mit Analogausgang
Titel japanisch
Transactions of the Information Processing Society of Japan ; 36 , 3 ; 627-635
1995
9 Seiten, 16 Quellen
Aufsatz (Zeitschrift)
Japanisch