Research on Time Performance Optimization of LSTM Model for Pedestrian Volume Prediction
10th IEEE Joint International Information Technology and Artificial Intelligence Conference, ITAIC 2022
; 2022-June:517-521, 2022.
Article
in English
| Scopus | ID: covidwho-2018926
ABSTRACT
The prediction of pedestrian volume in public area is of great significance to maintain personnel safety and improve the level of public management. Especially under the situation of COVID-19 prevention and control, the prediction of pedestrian volume within closed public spaces has a higher demand. While long short-term memory (LSTM), is often used to establish the prediction model of time series, for this purpose, taking the pedestrian flow prediction as the application background, the influence of the activation function on the time performance of LSTM model is studied, and an optimized scheme of the activation function, which can significantly improve the time performance while ensuring the prediction accuracy is proposed in this paper. The experimental results based on pedestrian flow prediction show that the time performance of the optimized LSTM model is improved by about 12.8% compared with the traditional model, and the prediction accuracy is even slightly increased. © 2022 IEEE.
activation function; deep learning; long short-term memory(LSTM); pedestrian volume prediction; Chemical activation; Forecasting; Human resource management; Pedestrian safety; Activation functions; Flow prediction; Memory modeling; On-time performance; Pedestrian flow; Performance; Prediction accuracy; Volume predictions; Long short-term memory
Full text:
Available
Collection:
Databases of international organizations
Database:
Scopus
Type of study:
Prognostic study
Language:
English
Journal:
10th IEEE Joint International Information Technology and Artificial Intelligence Conference, ITAIC 2022
Year:
2022
Document Type:
Article
Similar
MEDLINE
...
LILACS
LIS