The continuous increase in car sales parking problem is serious, depending on the date of the resolution and to resolve the difficulties of expanding the parking facilities for efficient operation and management of existing parking facilities through ...
The continuous increase in car sales parking problem is serious, depending on the date of the resolution and to resolve the difficulties of expanding the parking facilities for efficient operation and management of existing parking facilities through the park management plan will need to be taken.
Communications and computer technology to the development of the management applied to parking, parking management system, the introduction and implementation of effective alternative is emerging.
However, there need to be weighted more, whereas, in research and development level of the parking information system related to the incomplete research and development on a dramatic and daring package will continue to be said.
The system design required for parking information, parking information systems technology development of the prediction algorithm in the field to be the core of the system's development, as well as the development of algorithms to predict the actual test and verification is necessary to be involved. In addition to the existing system to upgrade the hardware progress and will be done.
Therefore, the parking information systems, perform research and development of the prediction algorithm, the actual test (Test Bed operation), perform a comprehensive test and verification of results.
Also, unlike the existing system of information collection and delivery system based on USN introduced, ubiquitous in the field of Transportation (u-Transportation) to consider progress in the environment were present in the system plan.
Survey data obtained through the Test-Bed based on analysis of problems in the system and algorithms. Review were, vehicle in-out and share patterns, changes in A ~ D of the homogeneous section is divided into 4 sections, and verification of validity, Distributed analysis in the analysis and classification of the homogeneous section of the review of statistical note.
Homogeneous section A(AM Peak, 07:00~09:00) to the sudden increase of the vehicles entering in the event of sudden changes in rate of occupancy.
Section B(PM Non-Peak, 10:00~17:00) to the high rate of occupation and low rate of Error section, section C(PM Peak, 18:00~21:00) of the rapid increase in the vehicle to the exit out in the event of significant fluctuations in rate of occupancy, section D(Non-Peak, 22:00~06:00) a rate of occupancy and rate of Error all the fluctuations of the number of low in-out was to analyze a small section.
In all cases the distinction between the homogenous section applicable in the parking lot of the objective and fully qualified to present the concept of time is the difficult part. However, most will be the same in the parking lot to consider, so it will be.
After reviewing the characteristics of homogenous section, the rapid change of the affinity between the number of vehicles A, C, the relatively high rate of Error analysis and prediction of time and will increase that percentage, the more increased. Therefore, to improve the accuracy of algorithms that consider the characteristics of homogenous section separated by the verification of the algorithm to be said.
In particular, the affinity between section A and C rate of Error for the improvement of the algorithm can be implemented, and improvements were needed.
Time prediction algorithm based on parking demand and the reliability of the information provided, the longer is slow. Within a maximum of 10 minutes parking demand that could reasonably be identified through analysis. In addition, the application of the past, Data must be set within the range to a maximum of 60 minutes was expected.
Parking demand for the validation of prediction algorithms to improve the existing algorithm and verification algorithm comparison between experiment was conducted using design of experiment.
Analysis techniques, and prediction of the algorithm defined in the existing units, and the prediction time is set by setting the number of scenarios, depending on the case of verification. Verification, the t-Test (P-Value), MAPE, RMSE was used.
Verification of the algorithm results with scenario section A, the affinity between exponential smoothing, section C, the Simple average method and exponential smoothing, sectionB and D the Simple average method t-Test values, MAPE, RMSE is the most minimal of analyzers corporation was analyzed.
Section A, the existing prediction algorithms, the value of α by changing the unit to apply the analysis to minimize rate of Error was. Section C of the case, exponential smoothing(α = 0.3) and Simple average method Alternately applied to the analysis that was to minimize rate of Error. Section B and D the exponential smoothing corporation than the simple Analyzers Simple average method to rate of Error to minimize the prediction was expected.
A prediction of the time interval increases to increase the weight of the exponential smoothing of the analysis was appropriate to apply, the most recent prediction in the past by using the Data to provide the exact value is. Therefore, the change in parking patterns in the recent past between the extreme predictions of Data to be using will have to implement the algorithm.